![\"Open](\"https://colab.research.google.com/assets/colab-badge.svg\"){kind=icon}
"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "hK-SsrOGZjBd",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "!pip install openreview-py\n",
+ "!pip install pipreqs"
+ ],
+ "execution_count": 0,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "QRo-6UJ2ZjBm",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "%matplotlib inline\n",
+ "\n",
+ "import os\n",
+ "import re\n",
+ "import sys\n",
+ "import requests\n",
+ "import openreview\n",
+ "import pandas as pd\n",
+ "import matplotlib.pyplot as plt\n",
+ "\n",
+ "from random import choice\n",
+ "from wordcloud import WordCloud\n",
+ "from urllib.parse import urlparse"
+ ],
+ "execution_count": 0,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "N7DCgCBEZjBw",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "0b119d2c-1196-4bfb-d4fc-3458e425ac2f"
+ },
+ "source": [
+ "client = openreview.Client(baseurl=\"https://openreview.net\")\n",
+ "\n",
+ "blind_notes = {\n",
+ " note.id: note\n",
+ " for note in openreview.tools.iterget_notes(\n",
+ " client,\n",
+ " invitation=\"ICLR.cc/2020/Conference/-/Blind_Submission\",\n",
+ " details=\"original\",\n",
+ " )\n",
+ "}\n",
+ "\n",
+ "all_decision_notes = openreview.tools.iterget_notes(\n",
+ " client, invitation=\"ICLR.cc/2020/Conference/Paper.*/-/Decision\"\n",
+ ")\n",
+ "\n",
+ "accepted_submissions = [\n",
+ " blind_notes[decision_note.forum]\n",
+ " for decision_note in all_decision_notes\n",
+ " if \"Accept\" in decision_note.content[\"decision\"]\n",
+ "]\n",
+ "\n",
+ "len(accepted_submissions)"
+ ],
+ "execution_count": 2,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "687"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 2
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "4VZY4_AXZjB3",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "code_present = 0\n",
+ "code_links = []\n",
+ "for note in accepted_submissions:\n",
+ " try:\n",
+ " code_links.append(note.content[\"code\"])\n",
+ " # print(\"code found\")\n",
+ " code_present += 1\n",
+ " except:\n",
+ " print(\"Unexpected error:\", sys.exc_info()[0])"
+ ],
+ "execution_count": 0,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "UZhxqcAPZjB8",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "b079815c-b786-42b7-e0b8-71f0252623cc"
+ },
+ "source": [
+ "code_present"
+ ],
+ "execution_count": 4,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "344"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 4
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "l3pZyZctZjCC",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 54
+ },
+ "outputId": "911be15a-2ec1-4bdd-f4ae-69aa79bb4fd9"
+ },
+ "source": [
+ "urlparse(choice(code_links))"
+ ],
+ "execution_count": 5,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "ParseResult(scheme='https', netloc='drive.google.com', path='/drive/folders/1kIOc4SlAJllUJsrr2OnZ4izIQIw2JexU', params='', query='usp=sharing', fragment='')"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 5
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "Wym6x80tZjCJ",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "code_links_df = pd.DataFrame({\"links\": code_links})"
+ ],
+ "execution_count": 0,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "12Is5bPeZjCN",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "code_links_df[\"domains\"] = code_links_df.links.apply(lambda x: urlparse(x)[1])"
+ ],
+ "execution_count": 0,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "54IYMqW_ZjCR",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 503
+ },
+ "outputId": "af66e02d-1009-4e54-fcf1-6c8da0dbb767"
+ },
+ "source": [
+ "code_links_df.domains.value_counts()"
+ ],
+ "execution_count": 8,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "github.com 268\n",
+ "drive.google.com 28\n",
+ "www.dropbox.com 9\n",
+ "anonymous.4open.science 8\n",
+ "bit.ly 5\n",
+ "storage.googleapis.com 3\n",
+ "s000.tinyupload.com 2\n",
+ "sites.google.com 2\n",
+ "docs.google.com 1\n",
+ "nikaashpuri.github.io 1\n",
+ "nitishgupta.github.io 1\n",
+ "clevrer.csail.mit.edu 1\n",
+ "wgrathwohl.github.io 1\n",
+ "dap.csail.mit.edu 1\n",
+ "www.robots.ox.ac.uk 1\n",
+ "www.daml.in.tum.de 1\n",
+ "automated-discovery.github.io 1\n",
+ "rohitgirdhar.github.io 1\n",
+ "danijar.com 1\n",
+ "www.github.com 1\n",
+ "anonfile.com 1\n",
+ "goo.gl 1\n",
+ "toiaydcdyywlhzvlob.github.io 1\n",
+ "whyu.me 1\n",
+ "www.sendspace.com 1\n",
+ "mega.nz 1\n",
+ "www.cs.cmu.edu 1\n",
+ "Name: domains, dtype: int64"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 8
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "fnF84fCIZjCW",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "temp_link = \"\"\n",
+ "\n",
+ "\n",
+ "def clean_github_link(link):\n",
+ " link = link.strip()\n",
+ " if not link[-4:] == \".git\":\n",
+ " return link + \".git\"\n",
+ " else:\n",
+ " return link\n",
+ "\n",
+ "\n",
+ "github_repo_links = (\n",
+ " code_links_df.loc[code_links_df.domains == \"github.com\"]\n",
+ " .links.apply(clean_github_link)\n",
+ " .values\n",
+ ")"
+ ],
+ "execution_count": 0,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "KokWVr08ZjCc",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 764
+ },
+ "outputId": "46858585-9ff6-4bd7-c7eb-a3e5ff61e7f0"
+ },
+ "source": [
+ "# takes about 24 minutes to download\n",
+ "for link in github_repo_links:\n",
+ " !git clone $link --depth 1 --quiet"
+ ],
+ "execution_count": 10,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "fatal: repository 'https://github.com/tensorflow/addons/blob/master/tensorflow_addons/optimizers/lamb.py.git/' not found\n",
+ "fatal: repository 'https://github.com/carloderamo/shared/tree/master.git/' not found\n",
+ "Cloning into 'proxsgd'...\n",
+ "remote: Enumerating objects: 59, done.\u001b[K\n",
+ "remote: Counting objects: 100% (59/59), done.\u001b[K\n",
+ "remote: Compressing objects: 100% (45/45), done.\u001b[K\n",
+ "remote: Total 59 (delta 19), reused 33 (delta 12), pack-reused 0\u001b[K\n",
+ "Unpacking objects: 100% (59/59), done.\n",
+ "/bin/bash: https://github.com/cc-hpc-itwm/proxsgd.git: No such file or directory\n",
+ "fatal: repository 'https://github.com/suraj-nair-1/google-research/tree/master/hierarchical_foresight.git/' not found\n",
+ "fatal: repository 'https://github.com/google-research/google-research/tree/master/cfq.git/' not found\n",
+ "remote: Not Found\n",
+ "fatal: repository 'https://github.com/hangg7/deformable-kernels/.git/' not found\n",
+ "fatal: repository 'https://github.com/google-research/google-research/tree/master/meta_learning_without_memorization.git/' not found\n",
+ "remote: Not Found\n",
+ "fatal: repository 'https://github.com/GRAM-nets.git/' not found\n",
+ "fatal: could not read Username for 'https://github.com': No such device or address\n",
+ "remote: Not Found\n",
+ "fatal: repository 'http://github.com/AvigdorZ.git/' not found\n",
+ "fatal: repository 'https://github.com/google-research/language/tree/master/language/bert_extraction.git/' not found\n",
+ "remote: Not Found\n",
+ "fatal: repository 'https://github.com/anonymous-sushi-armadillo.git/' not found\n",
+ "fatal: repository 'https://github.com/NeurEXT/NEXT-learning-to-plan/blob/master/main.ipynb.git/' not found\n",
+ "fatal: repository 'https://github.com/tensorflow/federated/tree/master/tensorflow_federated/python/research/gans.git/' not found\n",
+ "remote: Not Found\n",
+ "fatal: repository 'https://github.com/snap-stanford/pretrain-gnns/.git/' not found\n",
+ "remote: Not Found\n",
+ "fatal: repository 'https://github.com/PKU-AI-Edge/DGN/.git/' not found\n",
+ "fatal: repository 'https://github.com/google-research/google-research/tree/master/weak_disentangle.git/' not found\n",
+ "fatal: repository 'https://github.com/google/trax/tree/master/trax/models/reformer.git/' not found\n",
+ "Cloning into 'neural-tangent-kernel-UCI'...\n",
+ "remote: Enumerating objects: 38, done.\u001b[K\n",
+ "remote: Counting objects: 100% (38/38), done.\u001b[K\n",
+ "remote: Compressing objects: 100% (30/30), done.\u001b[K\n",
+ "remote: Total 38 (delta 16), reused 19 (delta 6), pack-reused 0\u001b[K\n",
+ "Unpacking objects: 100% (38/38), done.\n",
+ "/bin/bash: https://drive.google.com/open?id=1SdgWmhEcnm4qyaM9xrkN01VF9tj40WZS.git: No such file or directory\n",
+ "remote: Not Found\n",
+ "fatal: repository 'https://github.com/nathandelara/Spectral-Embedding-of-Regularized-Block-Models/.git/' not found\n",
+ "fatal: could not read Username for 'https://github.com': No such device or address\n",
+ "remote: Not Found\n",
+ "fatal: repository 'https://github.com/deepsphere.git/' not found\n",
+ "fatal: could not read Username for 'https://github.com': No such device or address\n"
+ ],
+ "name": "stdout"
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "mkyBz3bKZjCh",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 208
+ },
+ "outputId": "74c479f3-e31c-453a-f341-99ab4d5d1bb6"
+ },
+ "source": [
+ "code_links_df.loc[code_links_df.domains == \"github.com\"].links.apply(\n",
+ " lambda x: urlparse(x)[2].split(\"/\")[1]\n",
+ ").value_counts().head(10)"
+ ],
+ "execution_count": 11,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "google-research 10\n",
+ "facebookresearch 4\n",
+ "TAMU-VITA 3\n",
+ "TonghanWang 2\n",
+ "JHL-HUST 2\n",
+ "automl 2\n",
+ "epfml 2\n",
+ "eth-sri 2\n",
+ "haebeom-lee 2\n",
+ "tensorflow 2\n",
+ "Name: links, dtype: int64"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 11
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "4CLEHCpLZjCm",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "root = \".\"\n",
+ "dirlist = [item for item in os.listdir(root) if os.path.isdir(os.path.join(root, item))]\n",
+ "print(dirlist)"
+ ],
+ "execution_count": 0,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "U90rMdkjZjCs",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "dirlist.remove('.config')\n",
+ "dirlist.remove('sample_data')"
+ ],
+ "execution_count": 0,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "_seGdc8cZjCx",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "f8592b2b-9f92-42ee-eb92-6abb34870035"
+ },
+ "source": [
+ "len(dirlist)"
+ ],
+ "execution_count": 14,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "247"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 14
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "QLx3abB8ZjC2",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "# takes about 10 minutes to run\n",
+ "for repo in dirlist:\n",
+ " path = \"/content/\" + repo\n",
+ " if os.path.exists(path + \"/requirements.txt\"):\n",
+ " pass\n",
+ " else:\n",
+ " !pipreqs $path"
+ ],
+ "execution_count": 0,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "sLrWrQI0ZjC6",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "has_req_cnt = no_req_cnt = 0\n",
+ "for repo in dirlist:\n",
+ " path = \"/content/\" + repo\n",
+ " if os.path.exists(path + \"/requirements.txt\"):\n",
+ " has_req_cnt += 1\n",
+ " else:\n",
+ " no_req_cnt += 1"
+ ],
+ "execution_count": 0,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "ODsvh3LrZjC-",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "99a85edf-ad25-4c1b-e770-c43f879c03c8"
+ },
+ "source": [
+ "has_req_cnt, no_req_cnt"
+ ],
+ "execution_count": 17,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "(237, 10)"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 17
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "79lFy7-vZjDG",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 191
+ },
+ "outputId": "f610a4f9-c393-477a-bb77-3bed4114c899"
+ },
+ "source": [
+ "all_repo_names = []\n",
+ "all_tool_names = []\n",
+ "for repo in dirlist:\n",
+ " try:\n",
+ " repo_name = repo\n",
+ " with open(\"/content/\" + repo + \"/\" + \"requirements.txt\", \"r\") as f:\n",
+ " tools = f.readlines()\n",
+ " tool_names = \",\".join(tools).lower()\n",
+ "\n",
+ " all_repo_names.append(repo_name)\n",
+ " all_tool_names.append(tool_names)\n",
+ " except:\n",
+ " print(\"Unexpected error for \", repo, sys.exc_info()[0])"
+ ],
+ "execution_count": 18,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "Unexpected error for SI-NI-FGSM | \n", + " | all_repo_names | \n", + "all_tool_names | \n", + "
|---|---|---|
| 0 | \n", + "KP2D | \n", + "\\n | \n", + "
| 1 | \n", + "CN-DPM | \n", + "ipdb\\n,jupyterlab\\n,matplotlib\\n,numpy\\n,pyyam... | \n", + "
| 2 | \n", + "GLAD | \n", + "numpy==1.18.3\\n,pandas==1.0.3\\n,matplotlib==3.... | \n", + "
| 3 | \n", + "bert_score | \n", + "# pytorch\\n,torch>=1.0.0\\n,# progress bars in ... | \n", + "
| 4 | \n", + "delay_stability | \n", + "torchvision==0.6.0+cu101\\n,bokeh==1.4.0\\n,six=... | \n", + "
| \n", + " | all_repo_names | \n", + "all_tool_names | \n", + "
|---|---|---|
| 0 | \n", + "KP2D | \n", + "\\n | \n", + "
| 1 | \n", + "CN-DPM | \n", + "ipdb\\n,jupyterlab\\n,matplotlib\\n,numpy\\n,pyyam... | \n", + "
| 2 | \n", + "GLAD | \n", + "numpy==1.18.3\\n,pandas==1.0.3\\n,matplotlib==3.... | \n", + "
| 3 | \n", + "bert_score | \n", + "# pytorch\\n,torch>=1.0.0\\n,# progress bars in ... | \n", + "
| 4 | \n", + "delay_stability | \n", + "torchvision==0.6.0+cu101\\n,bokeh==1.4.0\\n,six=... | \n", + "
| \n", + " | all_repo_names | \n", + "all_tool_names | \n", + "all_tool_names_cleaned | \n", + "
|---|---|---|---|
| 0 | \n", + "KP2D | \n", + "\\n | \n", + "\n", + " |
| 1 | \n", + "CN-DPM | \n", + "ipdb\\n,jupyterlab\\n,matplotlib\\n,numpy\\n,pyyam... | \n", + "ipdb,jupyterlab,matplotlib,numpy,pyyaml,tensor... | \n", + "
| 2 | \n", + "GLAD | \n", + "numpy==1.18.3\\n,pandas==1.0.3\\n,matplotlib==3.... | \n", + "numpy,pandas,matplotlib,scipy,torch,networkx,s... | \n", + "
| 3 | \n", + "bert_score | \n", + "# pytorch\\n,torch>=1.0.0\\n,# progress bars in ... | \n", + "torch,tqdm,transformers,matplotlib,pandas,numpy | \n", + "
| 4 | \n", + "delay_stability | \n", + "torchvision==0.6.0+cu101\\n,bokeh==1.4.0\\n,six=... | \n", + "torchvision,bokeh,six,torch,numpy,scipy,pandas... | \n", + "
"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Improvements\n",
+ "\n",
+ "1. Tested on Google Colab\n",
+ "2. Updated the code to use data generators to process image data on the fly\n",
+ "3. Code style improved"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 1,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 52
+ },
+ "colab_type": "code",
+ "id": "JIuCtW728WD2",
+ "outputId": "c806e217-e60b-4e50-ef37-198106096fcd"
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "TensorFlow 1.x selected.\n"
+ ]
+ },
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "Using TensorFlow backend.\n"
+ ]
+ }
+ ],
+ "source": [
+ "%tensorflow_version 1.x\n",
+ "\n",
+ "import cv2\n",
+ "import math\n",
+ "import numpy as np\n",
+ "import tensorflow as tf\n",
+ "import keras.backend as K\n",
+ "\n",
+ "from keras.models import Model\n",
+ "from keras.optimizers import SGD\n",
+ "from keras.utils import np_utils\n",
+ "from keras.datasets import cifar10\n",
+ "from keras.layers.core import Layer\n",
+ "from keras.callbacks import LearningRateScheduler\n",
+ "from keras.initializers import glorot_uniform, Constant\n",
+ "from keras.preprocessing.image import ImageDataGenerator\n",
+ "from keras.layers import (\n",
+ " Conv2D,\n",
+ " MaxPool2D,\n",
+ " Dropout,\n",
+ " Dense,\n",
+ " Input,\n",
+ " concatenate,\n",
+ " GlobalAveragePooling2D,\n",
+ " AveragePooling2D,\n",
+ " Flatten,\n",
+ ")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {
+ "colab": {},
+ "colab_type": "code",
+ "id": "m5p4szZUEQeZ"
+ },
+ "outputs": [],
+ "source": [
+ "(X_train, y_train), (X_valid, y_valid) = cifar10.load_data()\n",
+ "y_train = np_utils.to_categorical(y_train, 10)\n",
+ "y_valid = np_utils.to_categorical(y_valid, 10)\n",
+ "\n",
+ "\n",
+ "def CustomImageDataGenerator(X, y, batch_size):\n",
+ " generator = ImageDataGenerator(rescale=1 / 255.0, dtype=\"float32\")\n",
+ "\n",
+ " datagen = generator.flow(X, y, batch_size=batch_size)\n",
+ "\n",
+ " while True:\n",
+ " X, y = datagen.__next__()\n",
+ " X = np.array([cv2.resize(img, (224, 224)) for img in X[:, :, :, :]])\n",
+ " yield X, [y, y, y]"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {
+ "colab": {},
+ "colab_type": "code",
+ "id": "0ZkGA5pA8XKu"
+ },
+ "outputs": [],
+ "source": [
+ "def inception_module(\n",
+ " x,\n",
+ " filters_1x1,\n",
+ " filters_3x3_reduce,\n",
+ " filters_3x3,\n",
+ " filters_5x5_reduce,\n",
+ " filters_5x5,\n",
+ " filters_pool_proj,\n",
+ " name=None,\n",
+ "):\n",
+ "\n",
+ " conv_1x1 = Conv2D(\n",
+ " filters=filters_1x1,\n",
+ " kernel_size=(1, 1),\n",
+ " padding=\"same\",\n",
+ " activation=\"relu\",\n",
+ " kernel_initializer=kernel_init,\n",
+ " bias_initializer=bias_init,\n",
+ " )(x)\n",
+ "\n",
+ " conv_3x3 = Conv2D(\n",
+ " filters=filters_3x3_reduce,\n",
+ " kernel_size=(1, 1),\n",
+ " padding=\"same\",\n",
+ " activation=\"relu\",\n",
+ " kernel_initializer=kernel_init,\n",
+ " bias_initializer=bias_init,\n",
+ " )(x)\n",
+ "\n",
+ " conv_3x3 = Conv2D(\n",
+ " filters=filters_3x3,\n",
+ " kernel_size=(3, 3),\n",
+ " padding=\"same\",\n",
+ " activation=\"relu\",\n",
+ " kernel_initializer=kernel_init,\n",
+ " bias_initializer=bias_init,\n",
+ " )(conv_3x3)\n",
+ "\n",
+ " conv_5x5 = Conv2D(\n",
+ " filters=filters_5x5_reduce,\n",
+ " kernel_size=(1, 1),\n",
+ " padding=\"same\",\n",
+ " activation=\"relu\",\n",
+ " kernel_initializer=kernel_init,\n",
+ " bias_initializer=bias_init,\n",
+ " )(x)\n",
+ "\n",
+ " conv_5x5 = Conv2D(\n",
+ " filters=filters_5x5,\n",
+ " kernel_size=(5, 5),\n",
+ " padding=\"same\",\n",
+ " activation=\"relu\",\n",
+ " kernel_initializer=kernel_init,\n",
+ " bias_initializer=bias_init,\n",
+ " )(conv_5x5)\n",
+ "\n",
+ " pool_proj = MaxPool2D(pool_size=(3, 3), strides=(1, 1), padding=\"same\")(x)\n",
+ "\n",
+ " pool_proj = Conv2D(\n",
+ " filters=filters_pool_proj,\n",
+ " kernel_size=(1, 1),\n",
+ " padding=\"same\",\n",
+ " activation=\"relu\",\n",
+ " kernel_initializer=kernel_init,\n",
+ " bias_initializer=bias_init,\n",
+ " )(pool_proj)\n",
+ "\n",
+ " output = concatenate(\n",
+ " inputs=[conv_1x1, conv_3x3, conv_5x5, pool_proj], axis=3, name=name\n",
+ " )\n",
+ "\n",
+ " return output"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {
+ "colab": {},
+ "colab_type": "code",
+ "id": "1y73xcMu8XNa"
+ },
+ "outputs": [],
+ "source": [
+ "kernel_init = glorot_uniform()\n",
+ "bias_init = Constant(value=0.2)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 5,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 332
+ },
+ "colab_type": "code",
+ "id": "7ZYsSUO88XPu",
+ "outputId": "a9035306-623d-419d-cf0e-819e72846bdc"
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:66: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:541: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4432: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4267: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4271: The name tf.nn.avg_pool is deprecated. Please use tf.nn.avg_pool2d instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:148: The name tf.placeholder_with_default is deprecated. Please use tf.compat.v1.placeholder_with_default instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3733: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
+ "Instructions for updating:\n",
+ "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
+ "WARNING:tensorflow:Large dropout rate: 0.7 (>0.5). In TensorFlow 2.x, dropout() uses dropout rate instead of keep_prob. Please ensure that this is intended.\n",
+ "WARNING:tensorflow:Large dropout rate: 0.7 (>0.5). In TensorFlow 2.x, dropout() uses dropout rate instead of keep_prob. Please ensure that this is intended.\n"
+ ]
+ }
+ ],
+ "source": [
+ "input_layer = Input(shape=(224, 224, 3))\n",
+ "\n",
+ "x = Conv2D(\n",
+ " 64,\n",
+ " (7, 7),\n",
+ " padding=\"same\",\n",
+ " strides=(2, 2),\n",
+ " activation=\"relu\",\n",
+ " name=\"conv_1_7x7/2\",\n",
+ " kernel_initializer=kernel_init,\n",
+ " bias_initializer=bias_init,\n",
+ ")(input_layer)\n",
+ "x = MaxPool2D((3, 3), padding=\"same\", strides=(2, 2), name=\"max_pool_1_3x3/2\")(x)\n",
+ "x = Conv2D(\n",
+ " 64, (1, 1), padding=\"same\", strides=(1, 1), activation=\"relu\", name=\"conv_2a_3x3/1\"\n",
+ ")(x)\n",
+ "x = Conv2D(\n",
+ " 192, (3, 3), padding=\"same\", strides=(1, 1), activation=\"relu\", name=\"conv_2b_3x3/1\"\n",
+ ")(x)\n",
+ "x = MaxPool2D((3, 3), padding=\"same\", strides=(2, 2), name=\"max_pool_2_3x3/2\")(x)\n",
+ "\n",
+ "x = inception_module(\n",
+ " x,\n",
+ " filters_1x1=64,\n",
+ " filters_3x3_reduce=96,\n",
+ " filters_3x3=128,\n",
+ " filters_5x5_reduce=16,\n",
+ " filters_5x5=32,\n",
+ " filters_pool_proj=32,\n",
+ " name=\"inception_3a\",\n",
+ ")\n",
+ "\n",
+ "x = inception_module(\n",
+ " x,\n",
+ " filters_1x1=128,\n",
+ " filters_3x3_reduce=128,\n",
+ " filters_3x3=192,\n",
+ " filters_5x5_reduce=32,\n",
+ " filters_5x5=96,\n",
+ " filters_pool_proj=64,\n",
+ " name=\"inception_3b\",\n",
+ ")\n",
+ "\n",
+ "x = MaxPool2D((3, 3), padding=\"same\", strides=(2, 2), name=\"max_pool_3_3x3/2\")(x)\n",
+ "\n",
+ "x = inception_module(\n",
+ " x,\n",
+ " filters_1x1=192,\n",
+ " filters_3x3_reduce=96,\n",
+ " filters_3x3=208,\n",
+ " filters_5x5_reduce=16,\n",
+ " filters_5x5=48,\n",
+ " filters_pool_proj=64,\n",
+ " name=\"inception_4a\",\n",
+ ")\n",
+ "\n",
+ "\n",
+ "x1 = AveragePooling2D((5, 5), strides=3)(x)\n",
+ "x1 = Conv2D(128, (1, 1), padding=\"same\", activation=\"relu\")(x1)\n",
+ "x1 = Flatten()(x1)\n",
+ "x1 = Dense(1024, activation=\"relu\")(x1)\n",
+ "x1 = Dropout(0.7)(x1)\n",
+ "x1 = Dense(10, activation=\"softmax\", name=\"auxilliary_output_1\")(x1)\n",
+ "\n",
+ "x = inception_module(\n",
+ " x,\n",
+ " filters_1x1=160,\n",
+ " filters_3x3_reduce=112,\n",
+ " filters_3x3=224,\n",
+ " filters_5x5_reduce=24,\n",
+ " filters_5x5=64,\n",
+ " filters_pool_proj=64,\n",
+ " name=\"inception_4b\",\n",
+ ")\n",
+ "\n",
+ "x = inception_module(\n",
+ " x,\n",
+ " filters_1x1=128,\n",
+ " filters_3x3_reduce=128,\n",
+ " filters_3x3=256,\n",
+ " filters_5x5_reduce=24,\n",
+ " filters_5x5=64,\n",
+ " filters_pool_proj=64,\n",
+ " name=\"inception_4c\",\n",
+ ")\n",
+ "\n",
+ "x = inception_module(\n",
+ " x,\n",
+ " filters_1x1=112,\n",
+ " filters_3x3_reduce=144,\n",
+ " filters_3x3=288,\n",
+ " filters_5x5_reduce=32,\n",
+ " filters_5x5=64,\n",
+ " filters_pool_proj=64,\n",
+ " name=\"inception_4d\",\n",
+ ")\n",
+ "\n",
+ "\n",
+ "x2 = AveragePooling2D((5, 5), strides=3)(x)\n",
+ "x2 = Conv2D(128, (1, 1), padding=\"same\", activation=\"relu\")(x2)\n",
+ "x2 = Flatten()(x2)\n",
+ "x2 = Dense(1024, activation=\"relu\")(x2)\n",
+ "x2 = Dropout(0.7)(x2)\n",
+ "x2 = Dense(10, activation=\"softmax\", name=\"auxilliary_output_2\")(x2)\n",
+ "\n",
+ "x = inception_module(\n",
+ " x,\n",
+ " filters_1x1=256,\n",
+ " filters_3x3_reduce=160,\n",
+ " filters_3x3=320,\n",
+ " filters_5x5_reduce=32,\n",
+ " filters_5x5=128,\n",
+ " filters_pool_proj=128,\n",
+ " name=\"inception_4e\",\n",
+ ")\n",
+ "\n",
+ "x = MaxPool2D((3, 3), padding=\"same\", strides=(2, 2), name=\"max_pool_4_3x3/2\")(x)\n",
+ "\n",
+ "x = inception_module(\n",
+ " x,\n",
+ " filters_1x1=256,\n",
+ " filters_3x3_reduce=160,\n",
+ " filters_3x3=320,\n",
+ " filters_5x5_reduce=32,\n",
+ " filters_5x5=128,\n",
+ " filters_pool_proj=128,\n",
+ " name=\"inception_5a\",\n",
+ ")\n",
+ "\n",
+ "x = inception_module(\n",
+ " x,\n",
+ " filters_1x1=384,\n",
+ " filters_3x3_reduce=192,\n",
+ " filters_3x3=384,\n",
+ " filters_5x5_reduce=48,\n",
+ " filters_5x5=128,\n",
+ " filters_pool_proj=128,\n",
+ " name=\"inception_5b\",\n",
+ ")\n",
+ "\n",
+ "x = GlobalAveragePooling2D(name=\"avg_pool_5_3x3/1\")(x)\n",
+ "\n",
+ "x = Dropout(0.4)(x)\n",
+ "\n",
+ "x = Dense(10, activation=\"softmax\", name=\"output\")(x)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {
+ "colab": {},
+ "colab_type": "code",
+ "id": "K5gv4hDA8XSA"
+ },
+ "outputs": [],
+ "source": [
+ "model = Model(input_layer, [x, x1, x2], name=\"inception_v1\")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {
+ "colab": {},
+ "colab_type": "code",
+ "id": "oDaDZJCE8XUO"
+ },
+ "outputs": [],
+ "source": [
+ "model.summary()"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 7,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 106
+ },
+ "colab_type": "code",
+ "id": "UUAV2emp8XWL",
+ "outputId": "ccd0abe1-7985-4edf-ba9d-56f47c1d7eed"
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:793: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3576: The name tf.log is deprecated. Please use tf.math.log instead.\n",
+ "\n"
+ ]
+ }
+ ],
+ "source": [
+ "epochs = 25\n",
+ "initial_lrate = 0.01\n",
+ "\n",
+ "\n",
+ "def decay(epoch, steps=100):\n",
+ " initial_lrate = 0.01\n",
+ " drop = 0.96\n",
+ " epochs_drop = 8\n",
+ " lrate = initial_lrate * math.pow(drop, math.floor((1 + epoch) / epochs_drop))\n",
+ " return lrate\n",
+ "\n",
+ "\n",
+ "sgd = SGD(lr=initial_lrate, momentum=0.9, nesterov=False)\n",
+ "\n",
+ "lr_sc = LearningRateScheduler(decay, verbose=1)\n",
+ "\n",
+ "model.compile(\n",
+ " loss=[\n",
+ " \"categorical_crossentropy\",\n",
+ " \"categorical_crossentropy\",\n",
+ " \"categorical_crossentropy\",\n",
+ " ],\n",
+ " loss_weights=[1, 0.3, 0.3],\n",
+ " optimizer=sgd,\n",
+ " metrics=[\"accuracy\"],\n",
+ ")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 610
+ },
+ "colab_type": "code",
+ "id": "FpVMejKG7TDI",
+ "outputId": "9745f1b1-ab2a-4366-ad74-2ea2195d63b2"
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "WARNING:tensorflow:From /tensorflow-1.15.2/python3.6/tensorflow_core/python/ops/math_grad.py:1424: where (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
+ "Instructions for updating:\n",
+ "Use tf.where in 2.0, which has the same broadcast rule as np.where\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1033: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1020: The name tf.assign is deprecated. Please use tf.compat.v1.assign instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3005: The name tf.Session is deprecated. Please use tf.compat.v1.Session instead.\n",
+ "\n",
+ "Epoch 1/25\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:190: The name tf.get_default_session is deprecated. Please use tf.compat.v1.get_default_session instead.\n",
+ "\n"
+ ]
+ },
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "/usr/local/lib/python3.6/dist-packages/keras/engine/training_generator.py:49: UserWarning: Using a generator with `use_multiprocessing=True` and multiple workers may duplicate your data. Please consider using the `keras.utils.Sequence class.\n",
+ " UserWarning('Using a generator with `use_multiprocessing=True`'\n"
+ ]
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:197: The name tf.ConfigProto is deprecated. Please use tf.compat.v1.ConfigProto instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:207: The name tf.global_variables is deprecated. Please use tf.compat.v1.global_variables instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:216: The name tf.is_variable_initialized is deprecated. Please use tf.compat.v1.is_variable_initialized instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:223: The name tf.variables_initializer is deprecated. Please use tf.compat.v1.variables_initializer instead.\n",
+ "\n",
+ "\n",
+ "Epoch 00001: LearningRateScheduler setting learning rate to 0.01.\n",
+ "200/200 [==============================] - 218s 1s/step - loss: 3.7005 - output_loss: 2.3221 - auxilliary_output_1_loss: 2.2912 - auxilliary_output_2_loss: 2.3033 - output_acc: 0.1143 - auxilliary_output_1_acc: 0.1238 - auxilliary_output_2_acc: 0.1117\n",
+ "Epoch 2/25\n",
+ "\n",
+ "Epoch 00002: LearningRateScheduler setting learning rate to 0.01.\n",
+ "200/200 [==============================] - 195s 974ms/step - loss: 3.2905 - output_loss: 2.0650 - auxilliary_output_1_loss: 2.0379 - auxilliary_output_2_loss: 2.0472 - output_acc: 0.2155 - auxilliary_output_1_acc: 0.2431 - auxilliary_output_2_acc: 0.2337\n",
+ "Epoch 3/25\n",
+ "\n",
+ "Epoch 00003: LearningRateScheduler setting learning rate to 0.01.\n",
+ " 48/200 [======>.......................] - ETA: 2:28 - loss: 3.1024 - output_loss: 1.9481 - auxilliary_output_1_loss: 1.9233 - auxilliary_output_2_loss: 1.9243 - output_acc: 0.2620 - auxilliary_output_1_acc: 0.2888 - auxilliary_output_2_acc: 0.2900"
+ ]
+ }
+ ],
+ "source": [
+ "model.fit_generator(\n",
+ " CustomImageDataGenerator(X_train, y_train, batch_size=256),\n",
+ " epochs=epochs,\n",
+ " steps_per_epoch=200,\n",
+ " use_multiprocessing=True,\n",
+ " workers=4,\n",
+ " callbacks=[lr_sc],\n",
+ ")"
+ ]
+ }
+ ],
+ "metadata": {
+ "accelerator": "GPU",
+ "colab": {
+ "collapsed_sections": [],
+ "include_colab_link": true,
+ "name": "Inception_v1_from_Scratch.ipynb",
+ "provenance": []
+ },
+ "kernelspec": {
+ "display_name": "Python 3",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.7.3"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}
diff --git a/Inception_From_Scratch/original_code/Inception_v1_from_Scratch.ipynb b/Inception_From_Scratch/original_code/Inception_v1_from_Scratch.ipynb
new file mode 100644
index 0000000..b0fce1b
--- /dev/null
+++ b/Inception_From_Scratch/original_code/Inception_v1_from_Scratch.ipynb
@@ -0,0 +1,321 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "### Implementation of GoogLeNet in Keras "
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "import keras\n",
+ "from keras.layers.core import Layer\n",
+ "import keras.backend as K\n",
+ "import tensorflow as tf\n",
+ "from keras.datasets import cifar10"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "from keras.models import Model\n",
+ "from keras.layers import Conv2D, MaxPool2D, \\\n",
+ " Dropout, Dense, Input, concatenate, \\\n",
+ " GlobalAveragePooling2D, AveragePooling2D,\\\n",
+ " Flatten\n",
+ "\n",
+ "import cv2 \n",
+ "import numpy as np \n",
+ "from keras.datasets import cifar10 \n",
+ "from keras import backend as K \n",
+ "from keras.utils import np_utils\n",
+ "\n",
+ "import math \n",
+ "from keras.optimizers import SGD \n",
+ "from keras.callbacks import LearningRateScheduler"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "num_classes = 10\n",
+ "\n",
+ "def load_cifar10_data(img_rows, img_cols):\n",
+ "\n",
+ " # Load cifar10 training and validation sets\n",
+ " (X_train, Y_train), (X_valid, Y_valid) = cifar10.load_data()\n",
+ "\n",
+ " # Resize training images\n",
+ " X_train = np.array([cv2.resize(img, (img_rows,img_cols)) for img in X_train[:,:,:,:]])\n",
+ " X_valid = np.array([cv2.resize(img, (img_rows,img_cols)) for img in X_valid[:,:,:,:]])\n",
+ "\n",
+ " # Transform targets to keras compatible format\n",
+ " Y_train = np_utils.to_categorical(Y_train, num_classes)\n",
+ " Y_valid = np_utils.to_categorical(Y_valid, num_classes)\n",
+ " \n",
+ " X_train = X_train.astype('float32')\n",
+ " X_valid = X_valid.astype('float32')\n",
+ "\n",
+ " # preprocess data\n",
+ " X_train = X_train / 255.0\n",
+ " X_valid = X_valid / 255.0\n",
+ "\n",
+ " return X_train, Y_train, X_valid, Y_valid"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "X_train, y_train, X_test, y_test = load_cifar10_data(224, 224)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "def inception_module(x,\n",
+ " filters_1x1,\n",
+ " filters_3x3_reduce,\n",
+ " filters_3x3,\n",
+ " filters_5x5_reduce,\n",
+ " filters_5x5,\n",
+ " filters_pool_proj,\n",
+ " name=None):\n",
+ " \n",
+ " conv_1x1 = Conv2D(filters_1x1, (1, 1), padding='same', activation='relu', kernel_initializer=kernel_init, bias_initializer=bias_init)(x)\n",
+ " \n",
+ " conv_3x3 = Conv2D(filters_3x3_reduce, (1, 1), padding='same', activation='relu', kernel_initializer=kernel_init, bias_initializer=bias_init)(x)\n",
+ " conv_3x3 = Conv2D(filters_3x3, (3, 3), padding='same', activation='relu', kernel_initializer=kernel_init, bias_initializer=bias_init)(conv_3x3)\n",
+ "\n",
+ " conv_5x5 = Conv2D(filters_5x5_reduce, (1, 1), padding='same', activation='relu', kernel_initializer=kernel_init, bias_initializer=bias_init)(x)\n",
+ " conv_5x5 = Conv2D(filters_5x5, (5, 5), padding='same', activation='relu', kernel_initializer=kernel_init, bias_initializer=bias_init)(conv_5x5)\n",
+ "\n",
+ " pool_proj = MaxPool2D((3, 3), strides=(1, 1), padding='same')(x)\n",
+ " pool_proj = Conv2D(filters_pool_proj, (1, 1), padding='same', activation='relu', kernel_initializer=kernel_init, bias_initializer=bias_init)(pool_proj)\n",
+ "\n",
+ " output = concatenate([conv_1x1, conv_3x3, conv_5x5, pool_proj], axis=3, name=name)\n",
+ " \n",
+ " return output"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "kernel_init = keras.initializers.glorot_uniform()\n",
+ "bias_init = keras.initializers.Constant(value=0.2)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "input_layer = Input(shape=(224, 224, 3))\n",
+ "\n",
+ "x = Conv2D(64, (7, 7), padding='same', strides=(2, 2), activation='relu', name='conv_1_7x7/2', kernel_initializer=kernel_init, bias_initializer=bias_init)(input_layer)\n",
+ "x = MaxPool2D((3, 3), padding='same', strides=(2, 2), name='max_pool_1_3x3/2')(x)\n",
+ "x = Conv2D(64, (1, 1), padding='same', strides=(1, 1), activation='relu', name='conv_2a_3x3/1')(x)\n",
+ "x = Conv2D(192, (3, 3), padding='same', strides=(1, 1), activation='relu', name='conv_2b_3x3/1')(x)\n",
+ "x = MaxPool2D((3, 3), padding='same', strides=(2, 2), name='max_pool_2_3x3/2')(x)\n",
+ "\n",
+ "x = inception_module(x,\n",
+ " filters_1x1=64,\n",
+ " filters_3x3_reduce=96,\n",
+ " filters_3x3=128,\n",
+ " filters_5x5_reduce=16,\n",
+ " filters_5x5=32,\n",
+ " filters_pool_proj=32,\n",
+ " name='inception_3a')\n",
+ "\n",
+ "x = inception_module(x,\n",
+ " filters_1x1=128,\n",
+ " filters_3x3_reduce=128,\n",
+ " filters_3x3=192,\n",
+ " filters_5x5_reduce=32,\n",
+ " filters_5x5=96,\n",
+ " filters_pool_proj=64,\n",
+ " name='inception_3b')\n",
+ "\n",
+ "x = MaxPool2D((3, 3), padding='same', strides=(2, 2), name='max_pool_3_3x3/2')(x)\n",
+ "\n",
+ "x = inception_module(x,\n",
+ " filters_1x1=192,\n",
+ " filters_3x3_reduce=96,\n",
+ " filters_3x3=208,\n",
+ " filters_5x5_reduce=16,\n",
+ " filters_5x5=48,\n",
+ " filters_pool_proj=64,\n",
+ " name='inception_4a')\n",
+ "\n",
+ "\n",
+ "x1 = AveragePooling2D((5, 5), strides=3)(x)\n",
+ "x1 = Conv2D(128, (1, 1), padding='same', activation='relu')(x1)\n",
+ "x1 = Flatten()(x1)\n",
+ "x1 = Dense(1024, activation='relu')(x1)\n",
+ "x1 = Dropout(0.7)(x1)\n",
+ "x1 = Dense(10, activation='softmax', name='auxilliary_output_1')(x1)\n",
+ "\n",
+ "x = inception_module(x,\n",
+ " filters_1x1=160,\n",
+ " filters_3x3_reduce=112,\n",
+ " filters_3x3=224,\n",
+ " filters_5x5_reduce=24,\n",
+ " filters_5x5=64,\n",
+ " filters_pool_proj=64,\n",
+ " name='inception_4b')\n",
+ "\n",
+ "x = inception_module(x,\n",
+ " filters_1x1=128,\n",
+ " filters_3x3_reduce=128,\n",
+ " filters_3x3=256,\n",
+ " filters_5x5_reduce=24,\n",
+ " filters_5x5=64,\n",
+ " filters_pool_proj=64,\n",
+ " name='inception_4c')\n",
+ "\n",
+ "x = inception_module(x,\n",
+ " filters_1x1=112,\n",
+ " filters_3x3_reduce=144,\n",
+ " filters_3x3=288,\n",
+ " filters_5x5_reduce=32,\n",
+ " filters_5x5=64,\n",
+ " filters_pool_proj=64,\n",
+ " name='inception_4d')\n",
+ "\n",
+ "\n",
+ "x2 = AveragePooling2D((5, 5), strides=3)(x)\n",
+ "x2 = Conv2D(128, (1, 1), padding='same', activation='relu')(x2)\n",
+ "x2 = Flatten()(x2)\n",
+ "x2 = Dense(1024, activation='relu')(x2)\n",
+ "x2 = Dropout(0.7)(x2)\n",
+ "x2 = Dense(10, activation='softmax', name='auxilliary_output_2')(x2)\n",
+ "\n",
+ "x = inception_module(x,\n",
+ " filters_1x1=256,\n",
+ " filters_3x3_reduce=160,\n",
+ " filters_3x3=320,\n",
+ " filters_5x5_reduce=32,\n",
+ " filters_5x5=128,\n",
+ " filters_pool_proj=128,\n",
+ " name='inception_4e')\n",
+ "\n",
+ "x = MaxPool2D((3, 3), padding='same', strides=(2, 2), name='max_pool_4_3x3/2')(x)\n",
+ "\n",
+ "x = inception_module(x,\n",
+ " filters_1x1=256,\n",
+ " filters_3x3_reduce=160,\n",
+ " filters_3x3=320,\n",
+ " filters_5x5_reduce=32,\n",
+ " filters_5x5=128,\n",
+ " filters_pool_proj=128,\n",
+ " name='inception_5a')\n",
+ "\n",
+ "x = inception_module(x,\n",
+ " filters_1x1=384,\n",
+ " filters_3x3_reduce=192,\n",
+ " filters_3x3=384,\n",
+ " filters_5x5_reduce=48,\n",
+ " filters_5x5=128,\n",
+ " filters_pool_proj=128,\n",
+ " name='inception_5b')\n",
+ "\n",
+ "x = GlobalAveragePooling2D(name='avg_pool_5_3x3/1')(x)\n",
+ "\n",
+ "x = Dropout(0.4)(x)\n",
+ "\n",
+ "x = Dense(10, activation='softmax', name='output')(x)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "model = Model(input_layer, [x, x1, x2], name='inception_v1')"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "model.summary()"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "epochs = 25\n",
+ "initial_lrate = 0.01\n",
+ "\n",
+ "def decay(epoch, steps=100):\n",
+ " initial_lrate = 0.01\n",
+ " drop = 0.96\n",
+ " epochs_drop = 8\n",
+ " lrate = initial_lrate * math.pow(drop, math.floor((1+epoch)/epochs_drop))\n",
+ " return lrate\n",
+ "\n",
+ "sgd = SGD(lr=initial_lrate, momentum=0.9, nesterov=False)\n",
+ "\n",
+ "lr_sc = LearningRateScheduler(decay, verbose=1)\n",
+ "\n",
+ "model.compile(loss=['categorical_crossentropy', 'categorical_crossentropy', 'categorical_crossentropy'], loss_weights=[1, 0.3, 0.3], optimizer=sgd, metrics=['accuracy'])"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "history = model.fit(X_train, [y_train, y_train, y_train], validation_data=(X_test, [y_test, y_test, y_test]), epochs=epochs, batch_size=256, callbacks=[lr_sc])"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.7.3"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}
diff --git a/NN_From_Scratch/README.md b/NN_From_Scratch/README.md
new file mode 100644
index 0000000..b75be7d
--- /dev/null
+++ b/NN_From_Scratch/README.md
@@ -0,0 +1,10 @@
+Understanding and coding Neural Networks From Scratch in Python and R
+===============================================================================
+
+This repository contains code for the article ["Understanding and coding Neural Networks From Scratch in Python and R"](https://www.analyticsvidhya.com/blog/2017/05/neural-network-from-scratch-in-python-and-r) article published on Analytics Vidhya
+
+Structure:
+---------
+
+- original_code/
+ - NN_From_Scratch_Python.ipynb
diff --git a/NN_From_Scratch/improvements/NN_From_Scratch_Python.ipynb b/NN_From_Scratch/improvements/NN_From_Scratch_Python.ipynb
new file mode 100644
index 0000000..8a46da1
--- /dev/null
+++ b/NN_From_Scratch/improvements/NN_From_Scratch_Python.ipynb
@@ -0,0 +1,1992 @@
+{
+ "nbformat": 4,
+ "nbformat_minor": 0,
+ "metadata": {
+ "accelerator": "GPU",
+ "colab": {
+ "name": "Neural Network from scratch using NumPy.ipynb",
+ "provenance": [],
+ "collapsed_sections": [],
+ "include_colab_link": true
+ },
+ "kernelspec": {
+ "display_name": "Python 3",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.7.3"
+ }
+ },
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "view-in-github",
+ "colab_type": "text"
+ },
+ "source": [
+ ""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "cEG1dhnuNFRk"
+ },
+ "source": [
+ "## Steps to build a Neural Network in NumPy\n",
+ "\n",
+ "> #### 1. Load the dataset \n",
+ "> #### 2. Define architecture of the model \n",
+ "> #### 3. Initialize the parameters\n",
+ "> #### 4. Implement forward propagation\n",
+ "> #### 5. Implement backward propagation\n",
+ "> #### 6. Train the model for multiple epochs \n",
+ "\n",
+ "---"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "sbgj7HfHNFRr"
+ },
+ "source": [
+ "### 1. Load the dataset "
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "E5S9HgBzNFRw",
+ "colab": {}
+ },
+ "source": [
+ "# importing required libraries\n",
+ "%matplotlib inline\n",
+ "\n",
+ "import numpy as np\n",
+ "import matplotlib.pyplot as plt"
+ ],
+ "execution_count": 1,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "ogs6CaXu2zeZ",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "2a129bea-6c3b-4eac-f1bb-2a6f07e167c1"
+ },
+ "source": [
+ "# version of numpy library\n",
+ "print(\"Version of numpy:\", np.__version__)"
+ ],
+ "execution_count": 2,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "Version of numpy: 1.18.5\n"
+ ],
+ "name": "stdout"
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "vNmvxGv723N6",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "aac2f147-a0c3-4232-f823-7ef8d4b109b3"
+ },
+ "source": [
+ "# version of matplotlib library\n",
+ "import matplotlib\n",
+ "\n",
+ "print(\"Version of matplotlib:\", matplotlib.__version__)"
+ ],
+ "execution_count": 3,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "Version of matplotlib: 3.2.2\n"
+ ],
+ "name": "stdout"
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "cf5NZ52KragQ",
+ "colab_type": "code",
+ "colab": {}
+ },
+ "source": [
+ "# set random seed\n",
+ "np.random.seed(42)"
+ ],
+ "execution_count": 4,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "H_h7HoPONFR_",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 121
+ },
+ "outputId": "77af76c4-e3fc-4a70-aca0-4cce73375d28"
+ },
+ "source": [
+ "# creating the input array\n",
+ "X = np.array([[1, 0, 0, 0], [1, 0, 1, 1], [0, 1, 0, 1]])\n",
+ "\n",
+ "print(\"Input:\\n\", X)\n",
+ "\n",
+ "# shape of input array\n",
+ "print(\"\\nShape of Input:\", X.shape)"
+ ],
+ "execution_count": 5,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "Input:\n",
+ " [[1 0 0 0]\n",
+ " [1 0 1 1]\n",
+ " [0 1 0 1]]\n",
+ "\n",
+ "Shape of Input: (3, 4)\n"
+ ],
+ "name": "stdout"
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "LVvQz5g39wo3",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 139
+ },
+ "outputId": "1f2e3a9e-2058-4943-dc25-b9e9816036c9"
+ },
+ "source": [
+ "# converting the input in matrix form\n",
+ "X = X.T\n",
+ "print(\"Input in matrix form:\\n\", X)\n",
+ "\n",
+ "# shape of input matrix\n",
+ "print(\"\\nShape of Input Matrix:\", X.shape)"
+ ],
+ "execution_count": 6,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "Input in matrix form:\n",
+ " [[1 1 0]\n",
+ " [0 0 1]\n",
+ " [0 1 0]\n",
+ " [0 1 1]]\n",
+ "\n",
+ "Shape of Input Matrix: (4, 3)\n"
+ ],
+ "name": "stdout"
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "IRe8JE0xNFSL",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 173
+ },
+ "outputId": "5cbd1d8b-e0ae-4505-ce48-cd3f8f1a6fd0"
+ },
+ "source": [
+ "# creating the output array\n",
+ "y = np.array([[1], [1], [0]])\n",
+ "\n",
+ "print(\"Actual Output:\\n\", y)\n",
+ "\n",
+ "# output in matrix form\n",
+ "y = y.T\n",
+ "\n",
+ "print(\"\\nOutput in matrix form:\\n\", y)\n",
+ "\n",
+ "# shape of input array\n",
+ "print(\"\\nShape of Output:\", y.shape)"
+ ],
+ "execution_count": 7,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "Actual Output:\n",
+ " [[1]\n",
+ " [1]\n",
+ " [0]]\n",
+ "\n",
+ "Output in matrix form:\n",
+ " [[1 1 0]]\n",
+ "\n",
+ "Shape of Output: (1, 3)\n"
+ ],
+ "name": "stdout"
+ }
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "tKf4Ji1-NFSV"
+ },
+ "source": [
+ "## 2. Define architecture of the model "
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "vlhBW0NNNFSg",
+ "colab": {}
+ },
+ "source": [
+ "inputLayer_neurons = X.shape[0] # number of features in data set\n",
+ "hiddenLayer_neurons = 3 # number of hidden layers neurons\n",
+ "outputLayer_neurons = 1 # number of neurons at output layer"
+ ],
+ "execution_count": 8,
+ "outputs": []
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "OoOsLucmNFSo"
+ },
+ "source": [
+ ""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "_Nmwj8RfNFSr"
+ },
+ "source": [
+ "## 3. Initialize the parameters\n",
+ "\n",
+ "NOTE: For simplicity, we are assuming that the bias for all the layers is 0"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "1T1IG-W8NFSu",
+ "colab": {}
+ },
+ "source": [
+ "# initializing weight\n",
+ "# Shape of weights_input_hidden should number of neurons at input layer * number of neurons at hidden layer\n",
+ "weights_input_hidden = np.random.uniform(size=(inputLayer_neurons, hiddenLayer_neurons))\n",
+ "\n",
+ "# Shape of weights_hidden_output should number of neurons at hidden layer * number of neurons at output layer\n",
+ "weights_hidden_output = np.random.uniform(\n",
+ " size=(hiddenLayer_neurons, outputLayer_neurons)\n",
+ ")"
+ ],
+ "execution_count": 9,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "Fpa1--9KNFS1",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "babbf8e8-90a5-49ff-a152-fa2dd2cc1eda"
+ },
+ "source": [
+ "# shape of weight matrix\n",
+ "weights_input_hidden.shape, weights_hidden_output.shape"
+ ],
+ "execution_count": 10,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "((4, 3), (3, 1))"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 10
+ }
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "srrDW1MNNFS-"
+ },
+ "source": [
+ "## 4. Implement forward propagation"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "sOcBji4iNFTE",
+ "colab": {}
+ },
+ "source": [
+ "# We are using sigmoid as an activation function so defining the sigmoid function here\n",
+ "\n",
+ "# defining the Sigmoid Function\n",
+ "def sigmoid(x):\n",
+ " return 1 / (1 + np.exp(-x))"
+ ],
+ "execution_count": 11,
+ "outputs": []
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "-g-SocwQNFTC"
+ },
+ "source": [
+ ""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "DO6AYHtGNFTM",
+ "colab": {}
+ },
+ "source": [
+ "# hidden layer activations\n",
+ "\n",
+ "hiddenLayer_linearTransform = np.dot(weights_input_hidden.T, X)\n",
+ "hiddenLayer_activations = sigmoid(hiddenLayer_linearTransform)"
+ ],
+ "execution_count": 12,
+ "outputs": []
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "o8zzYX6pNFTT"
+ },
+ "source": [
+ ""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "CuqKwiToNFTW",
+ "colab": {}
+ },
+ "source": [
+ "# calculating the output\n",
+ "outputLayer_linearTransform = np.dot(weights_hidden_output.T, hiddenLayer_activations)\n",
+ "output = sigmoid(outputLayer_linearTransform)"
+ ],
+ "execution_count": 13,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "BjPlMkVMNFTd",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "8a81d2de-a2f5-49c7-a6b5-bd65bd19cac3"
+ },
+ "source": [
+ "# output\n",
+ "output"
+ ],
+ "execution_count": 14,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "array([[0.68334694, 0.72697078, 0.71257368]])"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 14
+ }
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "mdFKMYyzNFTm"
+ },
+ "source": [
+ "## 5. Implement backward propagation"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "c2m3XBgZNFTn"
+ },
+ "source": [
+ ""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "IvUAAhlcNFTp",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "ef34ba99-13be-41f7-b886-ccaf9b80d403"
+ },
+ "source": [
+ "# calculating error\n",
+ "error = np.square(y - output) / 2\n",
+ "error"
+ ],
+ "execution_count": 15,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "array([[0.05013458, 0.03727248, 0.25388062]])"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 15
+ }
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "3H0vjBdNNFTw"
+ },
+ "source": [
+ "### Rate of change of error w.r.t weight between hidden and output layer"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "4cncCd1WNFTz"
+ },
+ "source": [
+ ""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "DqrhlDeDNFT1"
+ },
+ "source": [
+ "**a. Rate of change of error w.r.t output**\n",
+ "\n",
+ "**b. Rate of change of output w.r.t Z2**\n",
+ "\n",
+ "**c. Rate of change of Z2 w.r.t weights between hidden and output layer**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "bKdk5m4FNFT3",
+ "colab": {}
+ },
+ "source": [
+ "# rate of change of error w.r.t. output\n",
+ "error_wrt_output = -(y - output)"
+ ],
+ "execution_count": 16,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "Bl1PDwrBNFT9",
+ "colab": {}
+ },
+ "source": [
+ "# rate of change of output w.r.t. Z2\n",
+ "output_wrt_outputLayer_LinearTransform = np.multiply(output, (1 - output))"
+ ],
+ "execution_count": 17,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "3vLk1nxLNFUD",
+ "colab": {}
+ },
+ "source": [
+ "# rate of change of Z2 w.r.t. weights between hidden and output layer\n",
+ "outputLayer_LinearTransform_wrt_weights_hidden_output = hiddenLayer_activations"
+ ],
+ "execution_count": 18,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "UXXifY9QNFUI",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "525336f9-e81e-49ef-ac93-2d87baa7f29b"
+ },
+ "source": [
+ "# checking the shapes of partial derivatives\n",
+ "error_wrt_output.shape, output_wrt_outputLayer_LinearTransform.shape, outputLayer_LinearTransform_wrt_weights_hidden_output.shape"
+ ],
+ "execution_count": 19,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "((1, 3), (1, 3), (3, 3))"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 19
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "ZvtS7wCRNFUN",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "1eb93c65-713a-409d-cbc2-3f26ab201976"
+ },
+ "source": [
+ "# shape of weights of output layer\n",
+ "weights_hidden_output.shape"
+ ],
+ "execution_count": 20,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "(3, 1)"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 20
+ }
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "gC9zQEH6HON5"
+ },
+ "source": [
+ ""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "l3HNVYGONFUr",
+ "colab": {}
+ },
+ "source": [
+ "# rate of change of error w.r.t weight between hidden and output layer\n",
+ "error_wrt_weights_hidden_output = np.dot(\n",
+ " outputLayer_LinearTransform_wrt_weights_hidden_output,\n",
+ " (error_wrt_output * output_wrt_outputLayer_LinearTransform).T,\n",
+ ")"
+ ],
+ "execution_count": 21,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "cwyI1EGZNFUw",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "cd2a892b-4a5f-43eb-c543-3415ebd713bd"
+ },
+ "source": [
+ "error_wrt_weights_hidden_output.shape"
+ ],
+ "execution_count": 22,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "(3, 1)"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 22
+ }
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "sDFPg2SHNFU2"
+ },
+ "source": [
+ "### Rate of change of error w.r.t weight between input and hidden layer"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "_757MrjBNFU2"
+ },
+ "source": [
+ ""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "_nPYGXkeNFU4"
+ },
+ "source": [
+ "**a. Rate of change of error w.r.t output**\n",
+ "\n",
+ "**b. Rate of change of output w.r.t Z2**\n",
+ "\n",
+ "**c. Rate of change of Z2 w.r.t hidden layer activations**\n",
+ "\n",
+ "**d. Rate of change of hidden layer activations w.r.t Z1**\n",
+ "\n",
+ "**e. Rate of change of Z1 w.r.t weights between input and hidden layer**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "Sb7Ezxw9NFU6",
+ "colab": {}
+ },
+ "source": [
+ "# rate of change of error w.r.t. output\n",
+ "error_wrt_output = -(y - output)"
+ ],
+ "execution_count": 23,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "3-SGbNaoNFVA",
+ "colab": {}
+ },
+ "source": [
+ "# rate of change of output w.r.t. Z2\n",
+ "output_wrt_outputLayer_LinearTransform = np.multiply(output, (1 - output))"
+ ],
+ "execution_count": 24,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "amuoR7h6NFVF",
+ "colab": {}
+ },
+ "source": [
+ "# rate of change of Z2 w.r.t. hidden layer activations\n",
+ "outputLayer_LinearTransform_wrt_hiddenLayer_activations = weights_hidden_output"
+ ],
+ "execution_count": 25,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "YDUZEdWKNFVJ",
+ "colab": {}
+ },
+ "source": [
+ "# rate of change of hidden layer activations w.r.t. Z1\n",
+ "hiddenLayer_activations_wrt_hiddenLayer_linearTransform = np.multiply(\n",
+ " hiddenLayer_activations, (1 - hiddenLayer_activations)\n",
+ ")"
+ ],
+ "execution_count": 26,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "Ft4U6Td6NFVO",
+ "colab": {}
+ },
+ "source": [
+ "# rate of change of Z1 w.r.t. weights between input and hidden layer\n",
+ "hiddenLayer_linearTransform_wrt_weights_input_hidden = X"
+ ],
+ "execution_count": 27,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "A-hsfsi4NFVR",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "346c0c6a-5113-4f2b-eddb-cd0f48b02862"
+ },
+ "source": [
+ "# checking the shapes of partial derivatives\n",
+ "print(\n",
+ " error_wrt_output.shape,\n",
+ " output_wrt_outputLayer_LinearTransform.shape,\n",
+ " outputLayer_LinearTransform_wrt_hiddenLayer_activations.shape,\n",
+ " hiddenLayer_activations_wrt_hiddenLayer_linearTransform.shape,\n",
+ " hiddenLayer_linearTransform_wrt_weights_input_hidden.shape,\n",
+ ")"
+ ],
+ "execution_count": 28,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "(1, 3) (1, 3) (3, 1) (3, 3) (4, 3)\n"
+ ],
+ "name": "stdout"
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "1uka_yPrNFVV",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "0610b524-50cf-4d26-af18-0d11d5eeb5dd"
+ },
+ "source": [
+ "# shape of weights of hidden layer\n",
+ "weights_input_hidden.shape"
+ ],
+ "execution_count": 29,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "(4, 3)"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 29
+ }
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "gCeSm7vrHbHj"
+ },
+ "source": [
+ ""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "XTPNf3E5NFVs",
+ "colab": {}
+ },
+ "source": [
+ "# rate of change of error w.r.t weights between input and hidden layer\n",
+ "error_wrt_weights_input_hidden = np.dot(\n",
+ " hiddenLayer_linearTransform_wrt_weights_input_hidden,\n",
+ " (\n",
+ " hiddenLayer_activations_wrt_hiddenLayer_linearTransform\n",
+ " * np.dot(\n",
+ " outputLayer_LinearTransform_wrt_hiddenLayer_activations,\n",
+ " (output_wrt_outputLayer_LinearTransform * error_wrt_output),\n",
+ " )\n",
+ " ).T,\n",
+ ")"
+ ],
+ "execution_count": 30,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "_WN0I-mpNFVw",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 34
+ },
+ "outputId": "2f2c1dbe-52ba-47db-f2df-5e39d42aea9f"
+ },
+ "source": [
+ "error_wrt_weights_input_hidden.shape"
+ ],
+ "execution_count": 31,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "(4, 3)"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 31
+ }
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "W2bu4H5-NFVz"
+ },
+ "source": [
+ "### Update the parameters"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "-nmJnY_PNFV1"
+ },
+ "source": [
+ ""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "_r59xEpINFV2",
+ "colab": {}
+ },
+ "source": [
+ "# defining the learning rate\n",
+ "lr = 0.01"
+ ],
+ "execution_count": 32,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "aiBFNXd3NFV7",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 69
+ },
+ "outputId": "07b26520-3e50-423e-a218-80a68a875a9f"
+ },
+ "source": [
+ "# initial weights_hidden_output\n",
+ "weights_hidden_output"
+ ],
+ "execution_count": 33,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "array([[0.83244264],\n",
+ " [0.21233911],\n",
+ " [0.18182497]])"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 33
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "CuosFKUENFWB",
+ "scrolled": true,
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 86
+ },
+ "outputId": "b127e2f5-4d08-4d23-9f5f-aaf9971bc0af"
+ },
+ "source": [
+ "# initial weights_input_hidden\n",
+ "weights_input_hidden"
+ ],
+ "execution_count": 34,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "array([[0.37454012, 0.95071431, 0.73199394],\n",
+ " [0.59865848, 0.15601864, 0.15599452],\n",
+ " [0.05808361, 0.86617615, 0.60111501],\n",
+ " [0.70807258, 0.02058449, 0.96990985]])"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 34
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "D_Va2xywNFWF",
+ "colab": {}
+ },
+ "source": [
+ "# updating the weights of output layer\n",
+ "weights_hidden_output = weights_hidden_output - lr * error_wrt_weights_hidden_output"
+ ],
+ "execution_count": 35,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "ruFlc96BNFWL",
+ "colab": {}
+ },
+ "source": [
+ "# updating the weights of hidden layer\n",
+ "weights_input_hidden = weights_input_hidden - lr * error_wrt_weights_input_hidden"
+ ],
+ "execution_count": 36,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "NTf4nS1xNFWP",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 69
+ },
+ "outputId": "34890c4b-e441-4aa6-ce00-0b6b04a3a31d"
+ },
+ "source": [
+ "# updated weights_hidden_output\n",
+ "weights_hidden_output"
+ ],
+ "execution_count": 37,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "array([[0.83211079],\n",
+ " [0.21250681],\n",
+ " [0.18167831]])"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 37
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "7VYNPPNlNFWU",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 86
+ },
+ "outputId": "bd6bc8eb-4570-4a6f-ff28-347f99db88d6"
+ },
+ "source": [
+ "# updated weights_input_hidden\n",
+ "weights_input_hidden"
+ ],
+ "execution_count": 38,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "array([[0.37476062, 0.95075719, 0.7320294 ],\n",
+ " [0.59845481, 0.15594177, 0.15594545],\n",
+ " [0.05816641, 0.86618978, 0.60112315],\n",
+ " [0.70795169, 0.02052126, 0.96986892]])"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 38
+ }
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "SxLy6DZlNFWY"
+ },
+ "source": [
+ "## 6. Train the model for multiple epochs"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "8HKS9vIyNFWZ",
+ "colab": {}
+ },
+ "source": [
+ "# defining the model architecture\n",
+ "inputLayer_neurons = X.shape[0] # number of features in data set\n",
+ "hiddenLayer_neurons = 3 # number of hidden layers neurons\n",
+ "outputLayer_neurons = 1 # number of neurons at output layer\n",
+ "\n",
+ "# initializing weight\n",
+ "weights_input_hidden = np.random.uniform(size=(inputLayer_neurons, hiddenLayer_neurons))\n",
+ "weights_hidden_output = np.random.uniform(\n",
+ " size=(hiddenLayer_neurons, outputLayer_neurons)\n",
+ ")\n",
+ "\n",
+ "# defining the parameters\n",
+ "lr = 0.1\n",
+ "epochs = 1000"
+ ],
+ "execution_count": 39,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "_yVAcyW_NFWk",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 191
+ },
+ "outputId": "9e5d717f-e053-43c2-8da6-c2e437ed65de"
+ },
+ "source": [
+ "losses = []\n",
+ "for epoch in range(epochs):\n",
+ " ## Forward Propogation\n",
+ "\n",
+ " # calculating hidden layer activations\n",
+ " hiddenLayer_linearTransform = np.dot(weights_input_hidden.T, X)\n",
+ " hiddenLayer_activations = sigmoid(hiddenLayer_linearTransform)\n",
+ "\n",
+ " # calculating the output\n",
+ " outputLayer_linearTransform = np.dot(\n",
+ " weights_hidden_output.T, hiddenLayer_activations\n",
+ " )\n",
+ " output = sigmoid(outputLayer_linearTransform)\n",
+ "\n",
+ " ## Backward Propagation\n",
+ "\n",
+ " # calculating error\n",
+ " error = np.square(y - output) / 2\n",
+ "\n",
+ " # calculating rate of change of error w.r.t weight between hidden and output layer\n",
+ " error_wrt_output = -(y - output)\n",
+ " output_wrt_outputLayer_LinearTransform = np.multiply(output, (1 - output))\n",
+ " outputLayer_LinearTransform_wrt_weights_hidden_output = hiddenLayer_activations\n",
+ "\n",
+ " error_wrt_weights_hidden_output = np.dot(\n",
+ " outputLayer_LinearTransform_wrt_weights_hidden_output,\n",
+ " (error_wrt_output * output_wrt_outputLayer_LinearTransform).T,\n",
+ " )\n",
+ "\n",
+ " # calculating rate of change of error w.r.t weights between input and hidden layer\n",
+ " outputLayer_LinearTransform_wrt_hiddenLayer_activations = weights_hidden_output\n",
+ " hiddenLayer_activations_wrt_hiddenLayer_linearTransform = np.multiply(\n",
+ " hiddenLayer_activations, (1 - hiddenLayer_activations)\n",
+ " )\n",
+ " hiddenLayer_linearTransform_wrt_weights_input_hidden = X\n",
+ " error_wrt_weights_input_hidden = np.dot(\n",
+ " hiddenLayer_linearTransform_wrt_weights_input_hidden,\n",
+ " (\n",
+ " hiddenLayer_activations_wrt_hiddenLayer_linearTransform\n",
+ " * np.dot(\n",
+ " outputLayer_LinearTransform_wrt_hiddenLayer_activations,\n",
+ " (output_wrt_outputLayer_LinearTransform * error_wrt_output),\n",
+ " )\n",
+ " ).T,\n",
+ " )\n",
+ "\n",
+ " # updating the weights\n",
+ " weights_hidden_output = weights_hidden_output - lr * error_wrt_weights_hidden_output\n",
+ " weights_input_hidden = weights_input_hidden - lr * error_wrt_weights_input_hidden\n",
+ "\n",
+ " # print error at every 100th epoch\n",
+ " epoch_loss = np.average(error)\n",
+ " if epoch % 100 == 0:\n",
+ " print(f\"Error at epoch {epoch} is {epoch_loss:.5f}\")\n",
+ "\n",
+ " # appending the error of each epoch\n",
+ " losses.append(epoch_loss)"
+ ],
+ "execution_count": 40,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "Error at epoch 0 is 0.11553\n",
+ "Error at epoch 100 is 0.11082\n",
+ "Error at epoch 200 is 0.10606\n",
+ "Error at epoch 300 is 0.09845\n",
+ "Error at epoch 400 is 0.08483\n",
+ "Error at epoch 500 is 0.06396\n",
+ "Error at epoch 600 is 0.04206\n",
+ "Error at epoch 700 is 0.02641\n",
+ "Error at epoch 800 is 0.01719\n",
+ "Error at epoch 900 is 0.01190\n"
+ ],
+ "name": "stdout"
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "Ra5mTgwUNFWo",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 86
+ },
+ "outputId": "a7d314be-c563-4218-a191-40f0e5992773"
+ },
+ "source": [
+ "# updated w_ih\n",
+ "weights_input_hidden"
+ ],
+ "execution_count": 41,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "array([[ 1.25679149, 1.72312858, -0.27336634],\n",
+ " [-1.07615756, -1.73777864, 1.42316207],\n",
+ " [ 0.63053865, 0.88090942, -0.03448117],\n",
+ " [-0.56098781, -0.65506704, 0.61013995]])"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 41
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "ncRRRhdirair",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 69
+ },
+ "outputId": "17bef20e-894f-414e-ba0e-895f35f7d21e"
+ },
+ "source": [
+ "# updated w_ho\n",
+ "weights_hidden_output"
+ ],
+ "execution_count": 42,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "array([[ 1.45176252],\n",
+ " [ 2.59109536],\n",
+ " [-2.18347501]])"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "execution_count": 42
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "colab_type": "code",
+ "id": "WeN2dcc0NFW8",
+ "scrolled": true,
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 285
+ },
+ "outputId": "36c2519b-7b61-440f-99b1-8c476ba90415"
+ },
+ "source": [
+ "# visualizing the error after each epoch\n",
+ "plt.plot(np.arange(1, epochs + 1), np.array(losses))"
+ ],
+ "execution_count": 43,
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "["
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "Wizi4SPlG77Z",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 104
+ },
+ "outputId": "e922c653-cee0-4eb0-af7c-7b0d9cb70b56"
+ },
+ "source": [
+ "!pip install https://github.com/OlafenwaMoses/ImageAI/releases/download/2.0.1/imageai-2.0.1-py3-none-any.whl"
+ ],
+ "execution_count": 1,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "Collecting imageai==2.0.1\n",
+ "\u001b[?25l Downloading https://github.com/OlafenwaMoses/ImageAI/releases/download/2.0.1/imageai-2.0.1-py3-none-any.whl (137kB)\n",
+ "\u001b[K |████████████████████████████████| 143kB 819kB/s \n",
+ "\u001b[?25hInstalling collected packages: imageai\n",
+ "Successfully installed imageai-2.0.1\n"
+ ],
+ "name": "stdout"
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "UyNfH59fHK27",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 506
+ },
+ "outputId": "3b98d9f9-03de-4d27-867a-e299827f1cbf"
+ },
+ "source": [
+ "!wget https://github.com/OlafenwaMoses/ImageAI/releases/download/1.0/resnet50_coco_best_v2.0.1.h5\n",
+ "!wget https://cdn.analyticsvidhya.com/wp-content/uploads/2018/06/I1_2009_09_08_drive_0012_001351-768x223.png\n",
+ "\n",
+ "!mv I1_2009_09_08_drive_0012_001351-768x223.png image.png"
+ ],
+ "execution_count": 2,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "--2020-04-27 14:56:51-- https://github.com/OlafenwaMoses/ImageAI/releases/download/1.0/resnet50_coco_best_v2.0.1.h5\n",
+ "Resolving github.com (github.com)... 140.82.112.3\n",
+ "Connecting to github.com (github.com)|140.82.112.3|:443... connected.\n",
+ "HTTP request sent, awaiting response... 302 Found\n",
+ "Location: https://github-production-release-asset-2e65be.s3.amazonaws.com/125932201/e7ab678c-6146-11e8-85cc-26bc1cd06ab0?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20200427%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20200427T145651Z&X-Amz-Expires=300&X-Amz-Signature=944991495826ac4faa0676f53dedf6b9b34027551909a1c2cfdd159b09d5583f&X-Amz-SignedHeaders=host&actor_id=0&repo_id=125932201&response-content-disposition=attachment%3B%20filename%3Dresnet50_coco_best_v2.0.1.h5&response-content-type=application%2Foctet-stream [following]\n",
+ "--2020-04-27 14:56:51-- https://github-production-release-asset-2e65be.s3.amazonaws.com/125932201/e7ab678c-6146-11e8-85cc-26bc1cd06ab0?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20200427%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20200427T145651Z&X-Amz-Expires=300&X-Amz-Signature=944991495826ac4faa0676f53dedf6b9b34027551909a1c2cfdd159b09d5583f&X-Amz-SignedHeaders=host&actor_id=0&repo_id=125932201&response-content-disposition=attachment%3B%20filename%3Dresnet50_coco_best_v2.0.1.h5&response-content-type=application%2Foctet-stream\n",
+ "Resolving github-production-release-asset-2e65be.s3.amazonaws.com (github-production-release-asset-2e65be.s3.amazonaws.com)... 52.216.179.163\n",
+ "Connecting to github-production-release-asset-2e65be.s3.amazonaws.com (github-production-release-asset-2e65be.s3.amazonaws.com)|52.216.179.163|:443... connected.\n",
+ "HTTP request sent, awaiting response... 200 OK\n",
+ "Length: 152661008 (146M) [application/octet-stream]\n",
+ "Saving to: ‘resnet50_coco_best_v2.0.1.h5’\n",
+ "\n",
+ "resnet50_coco_best_ 100%[===================>] 145.59M 44.5MB/s in 3.5s \n",
+ "\n",
+ "2020-04-27 14:56:55 (42.1 MB/s) - ‘resnet50_coco_best_v2.0.1.h5’ saved [152661008/152661008]\n",
+ "\n",
+ "--2020-04-27 14:56:57-- https://cdn.analyticsvidhya.com/wp-content/uploads/2018/06/I1_2009_09_08_drive_0012_001351-768x223.png\n",
+ "Resolving cdn.analyticsvidhya.com (cdn.analyticsvidhya.com)... 104.26.15.185, 104.26.14.185, 2606:4700:20::681a:fb9, ...\n",
+ "Connecting to cdn.analyticsvidhya.com (cdn.analyticsvidhya.com)|104.26.15.185|:443... connected.\n",
+ "HTTP request sent, awaiting response... 200 OK\n",
+ "Length: 186458 (182K) [image/png]\n",
+ "Saving to: ‘I1_2009_09_08_drive_0012_001351-768x223.png’\n",
+ "\n",
+ "I1_2009_09_08_drive 100%[===================>] 182.09K --.-KB/s in 0.02s \n",
+ "\n",
+ "2020-04-27 14:56:59 (11.0 MB/s) - ‘I1_2009_09_08_drive_0012_001351-768x223.png’ saved [186458/186458]\n",
+ "\n"
+ ],
+ "name": "stdout"
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "Q-cKlNMEHVIt",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 52
+ },
+ "outputId": "838ab874-62a3-43b6-958f-c15cdb791ce0"
+ },
+ "source": [
+ "%tensorflow_version 1.x\n",
+ "from imageai.Detection import ObjectDetection\n",
+ "import os"
+ ],
+ "execution_count": 3,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "TensorFlow 1.x selected.\n"
+ ],
+ "name": "stdout"
+ },
+ {
+ "output_type": "stream",
+ "text": [
+ "Using TensorFlow backend.\n"
+ ],
+ "name": "stderr"
+ }
+ ]
+ },
+ {
+ "cell_type": "code",
+ "metadata": {
+ "id": "_CV0F896GjFV",
+ "colab_type": "code",
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 402
+ },
+ "outputId": "b140958c-717e-4cde-df78-b09f0e02c821"
+ },
+ "source": [
+ "execution_path = os.getcwd()\n",
+ "\n",
+ "detector = ObjectDetection()\n",
+ "detector.setModelTypeAsRetinaNet()\n",
+ "detector.setModelPath( os.path.join(execution_path , \"resnet50_coco_best_v2.0.1.h5\"))\n",
+ "detector.loadModel()\n",
+ "custom_objects = detector.CustomObjects(person=True, car=False)\n",
+ "detections = detector.detectCustomObjectsFromImage(input_image=os.path.join(execution_path , \"image.png\"), output_image_path=os.path.join(execution_path , \"updated_image.png\"), custom_objects=custom_objects, minimum_percentage_probability=65)\n",
+ "\n",
+ "\n",
+ "for eachObject in detections:\n",
+ " print(eachObject[\"name\"] + \" : \" + eachObject[\"percentage_probability\"] )\n",
+ " print(\"--------------------------------\")"
+ ],
+ "execution_count": 4,
+ "outputs": [
+ {
+ "output_type": "stream",
+ "text": [
+ "WARNING:tensorflow:From /tensorflow-1.15.2/python3.6/tensorflow_core/python/ops/resource_variable_ops.py:1630: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.\n",
+ "Instructions for updating:\n",
+ "If using Keras pass *_constraint arguments to layers.\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4070: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n",
+ "\n",
+ "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/imageai/Detection/keras_retinanet/backend/tensorflow_backend.py:22: The name tf.image.resize_images is deprecated. Please use tf.image.resize instead.\n",
+ "\n",
+ "tracking | \n", + " | filename | \n", + "label | \n", + "
|---|---|---|
| 0 | \n", + "0.png | \n", + "4 | \n", + "
| 1 | \n", + "1.png | \n", + "9 | \n", + "
| 2 | \n", + "2.png | \n", + "1 | \n", + "
| 3 | \n", + "3.png | \n", + "7 | \n", + "
| 4 | \n", + "4.png | \n", + "3 | \n", + "