From ca4197d87f98e02e11d948adcc6a1f6b0232fac7 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@outlook.com>
Date: Wed, 4 Sep 2019 19:37:08 +0800
Subject: [PATCH 01/22] Add GIOU loss

---
 tensorflow_addons/losses/BUILD             |  14 +++
 tensorflow_addons/losses/__init__.py       |   1 +
 tensorflow_addons/losses/giou_loss.py      | 138 +++++++++++++++++++++
 tensorflow_addons/losses/giou_loss_test.py |  20 +++
 4 files changed, 173 insertions(+)
 create mode 100644 tensorflow_addons/losses/giou_loss.py
 create mode 100644 tensorflow_addons/losses/giou_loss_test.py

diff --git a/tensorflow_addons/losses/BUILD b/tensorflow_addons/losses/BUILD
index 927f821fe2..c97351f0e4 100644
--- a/tensorflow_addons/losses/BUILD
+++ b/tensorflow_addons/losses/BUILD
@@ -8,6 +8,7 @@ py_library(
         "__init__.py",
         "contrastive.py",
         "focal_loss.py",
+        "giou_loss.py",
         "lifted.py",
         "metric_learning.py",
         "npairs.py",
@@ -47,6 +48,19 @@ py_test(
     ],
 )
 
+py_test(
+    name = "giou_loss_test",
+    size = "small",
+    srcs = [
+        "giou_loss_test.py",
+    ],
+    main = "giou_loss_test.py",
+    srcs_version = "PY2AND3",
+    deps = [
+        ":losses",
+    ],
+)
+
 py_test(
     name = "npairs_test",
     size = "small",
diff --git a/tensorflow_addons/losses/__init__.py b/tensorflow_addons/losses/__init__.py
index ff8e5094fa..efb00ec7a6 100644
--- a/tensorflow_addons/losses/__init__.py
+++ b/tensorflow_addons/losses/__init__.py
@@ -24,3 +24,4 @@
 from tensorflow_addons.losses.npairs import npairs_loss, NpairsLoss, npairs_multilabel_loss, NpairsMultilabelLoss
 from tensorflow_addons.losses.sparsemax_loss import sparsemax_loss, SparsemaxLoss
 from tensorflow_addons.losses.triplet import triplet_semihard_loss, TripletSemiHardLoss
+from tensorflow_addons.losses.giou_loss import giou_loss, GIOULoss
diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
new file mode 100644
index 0000000000..0abd713771
--- /dev/null
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -0,0 +1,138 @@
+# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Implements GIOU loss."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import tensorflow as tf
+from tensorflow_addons.utils import keras_utils
+
+
+@keras_utils.register_keras_custom_object
+class GIOULoss(tf.keras.losses.Loss):
+    """Implements the GIOU loss function.
+
+    GIOU loss was first introduced in the Generalized Intersection over Union paper
+    (https://giou.stanford.edu/GIoU.pdf). GIOU loss is a patch for model which IOU
+    is used in object detection.
+
+    Usage:
+
+    ```python
+    fl = tfa.losses.GIOU()
+    loss = fl(
+      [[0.97], [0.91], [0.03]],
+      [[1], [1], [0])
+    print('Loss: ', loss.numpy())  # Loss: [[0.03045921]
+                                            [0.09431068]
+                                            [0.31471074]
+    ```
+    Usage with tf.keras API:
+
+    ```python
+    model = tf.keras.Model(inputs, outputs)
+    model.compile('sgd', loss=tf.keras.losses.GIOULoss())
+    ```
+
+    Args
+      name: Op name
+
+    Returns:
+      Weighted loss float `Tensor`. If `reduction` is `NONE`, this has the same
+          shape as `y_true`; otherwise, it is scalar.
+
+    Raises:
+        ValueError: If the shape of `sample_weight` is invalid or value of
+          `gamma` is less than zero
+    """
+
+    def __init__(self,
+                 reduction=tf.keras.losses.Reduction.NONE,
+                 name='giou_loss'):
+        super(GIOULoss, self).__init__(name=name, reduction=reduction)
+
+    def call(self, y_true, y_pred):
+        return giou_loss(y_true, y_pred)
+
+    def get_config(self):
+        base_config = super(GIOULoss, self).get_config()
+        return base_config
+
+
+@keras_utils.register_keras_custom_object
+def giou_loss(y_true, y_pred):
+    """
+    Args
+        y_true: true targets tensor.
+        y_pred: predictions tensor.
+    
+    Returns:
+        Weighted loss float `Tensor`.
+    """
+
+    y_pred = tf.convert_to_tensor(y_pred)
+    y_true = tf.cast(y_true, y_pred.dtype)
+    giou = _giou(y_pred, y_true)
+
+    # compute the final loss and return
+    return 1 - giou
+
+
+@tf.function
+def _giou(b1, b2):
+    b1_ymin = tf.minimum(b1[:, 0], b1[:, 2])
+    b1_xmin = tf.minimum(b1[:, 1], b1[:, 3])
+    b1_ymax = tf.maximum(b1[:, 0], b1[:, 2])
+    b1_xmax = tf.maximum(b1[:, 1], b1[:, 3])
+    b2_ymin = tf.minimum(b2[:, 0], b2[:, 2])
+    b2_xmin = tf.minimum(b2[:, 1], b2[:, 3])
+    b2_ymax = tf.maximum(b2[:, 0], b2[:, 2])
+    b2_xmax = tf.maximum(b2[:, 1], b2[:, 3])
+    b1_area = (b1_ymax - b1_ymin) * (b1_xmax - b1_xmin)
+    b2_area = (b2_ymax - b2_ymin) * (b2_xmax - b1_xmin)
+    illegal_area_indexes = tf.cast(
+        tf.where(tf.logical_or(b1_area < 0, b2_area < 0)), tf.int32)
+    valid_area_indexes = tf.cast(
+        tf.where(tf.logical_and(b1_area >= 0, b2_area >= 0)), tf.int32)
+    # if b1_area<0 or b2_area<0:
+    #     return 0
+
+    intersect_ymin = tf.maximum(b1_ymin, b2_ymin)
+    intersect_xmin = tf.maximum(b1_xmin, b2_xmin)
+    intersect_ymax = tf.minimum(b1_ymax, b2_ymax)
+    intersect_xmax = tf.minimum(b1_xmax, b2_xmax)
+    intersect_area = tf.maximum(0,
+                                intersect_ymax - intersect_ymin) * tf.maximum(
+                                    0, intersect_xmax - intersect_xmin)
+
+    union_area = b1_area + b2_area - intersect_area
+    iou = intersect_area / union_area
+    indices = [valid_area_indexes, illegal_area_indexes]
+    data = [
+        tf.gather(iou, valid_area_indexes),
+        tf.zeros([tf.shape(illegal_area_indexes)[0], 1], tf.float64)
+    ]
+    iou = tf.dynamic_stitch(indices, data)
+    bc_ymin = tf.minimum(b1_ymin, b2_ymin)
+    bc_xmin = tf.minimum(b1_xmin, b2_xmin)
+    bc_ymax = tf.maximum(b1_ymax, b2_ymax)
+    bc_xmax = tf.maximum(b1_xmax, b2_xmax)
+
+    enclose_area = tf.maximum(0, bc_ymax - bc_ymin) * tf.maximum(
+        0, bc_xmax - bc_xmin)
+    giou = iou - (enclose_area - union_area) / enclose_area
+    return giou
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
new file mode 100644
index 0000000000..b53b70162a
--- /dev/null
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -0,0 +1,20 @@
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import tensorflow as tf
+from tensorflow_addons.utils import test_utils
+from tensorflow_addons.losses import giou_loss, GIOULoss
+
+
+@test_utils.run_all_in_graph_and_eager_modes
+class GIOULossTest(tf.test.TestCase):
+    def test_giou_loss(self):
+        pred_data = [[0, 0, 1, 1], [0, 0, 2, 2]]
+        true_data = [[0, 0, 1, 1], [0, 0, 1, 2]]
+        loss = giou_loss(true_data, pred_data)
+        self.assertAllClose(loss, tf.constant([0, 0.5], tf.float64))
+
+
+if __name__ == '__main__':
+    tf.test.main()

From 27ae8ead7f593f36b12f6afe94cbf8c2292462e9 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@outlook.com>
Date: Thu, 5 Sep 2019 10:46:07 +0800
Subject: [PATCH 02/22] Refact giou calculate

---
 tensorflow_addons/losses/giou_loss.py      | 46 +++++++++++++++-------
 tensorflow_addons/losses/giou_loss_test.py | 28 +++++++++++--
 2 files changed, 56 insertions(+), 18 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index 0abd713771..e4990ada48 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -27,8 +27,8 @@ class GIOULoss(tf.keras.losses.Loss):
     """Implements the GIOU loss function.
 
     GIOU loss was first introduced in the Generalized Intersection over Union paper
-    (https://giou.stanford.edu/GIoU.pdf). GIOU loss is a patch for model which IOU
-    is used in object detection.
+    (https://giou.stanford.edu/GIoU.pdf). GIOU is a enhance for model which use IOU
+    in object detection.
 
     Usage:
 
@@ -52,12 +52,7 @@ class GIOULoss(tf.keras.losses.Loss):
       name: Op name
 
     Returns:
-      Weighted loss float `Tensor`. If `reduction` is `NONE`, this has the same
-          shape as `y_true`; otherwise, it is scalar.
-
-    Raises:
-        ValueError: If the shape of `sample_weight` is invalid or value of
-          `gamma` is less than zero
+      GIOU loss float `Tensor`.
     """
 
     def __init__(self,
@@ -81,19 +76,39 @@ def giou_loss(y_true, y_pred):
         y_pred: predictions tensor.
     
     Returns:
-        Weighted loss float `Tensor`.
+        GIOU loss float `Tensor`.
     """
 
     y_pred = tf.convert_to_tensor(y_pred)
     y_true = tf.cast(y_true, y_pred.dtype)
-    giou = _giou(y_pred, y_true)
+    giou = giou_calculate(y_pred, y_true)
 
     # compute the final loss and return
     return 1 - giou
 
 
-@tf.function
-def _giou(b1, b2):
+def giou_calculate(b1, b2):
+    """
+    Args
+        b1: bbox.
+        b1: the other bbox.
+
+    Returns:
+        iou float `Tensor`, between [-1, 1].
+    """
+    return do_iou_calculate(b1, b2, mode='giou')
+
+
+def do_iou_calculate(b1, b2, mode='iou'):
+    """
+    Args
+        b1: bbox.
+        b1: the other bbox.
+        mode: one of ['iou', 'giou']
+
+    Returns:
+        iou float `Tensor`.
+    """
     b1_ymin = tf.minimum(b1[:, 0], b1[:, 2])
     b1_xmin = tf.minimum(b1[:, 1], b1[:, 3])
     b1_ymax = tf.maximum(b1[:, 0], b1[:, 2])
@@ -108,8 +123,6 @@ def _giou(b1, b2):
         tf.where(tf.logical_or(b1_area < 0, b2_area < 0)), tf.int32)
     valid_area_indexes = tf.cast(
         tf.where(tf.logical_and(b1_area >= 0, b2_area >= 0)), tf.int32)
-    # if b1_area<0 or b2_area<0:
-    #     return 0
 
     intersect_ymin = tf.maximum(b1_ymin, b2_ymin)
     intersect_xmin = tf.maximum(b1_xmin, b2_xmin)
@@ -127,6 +140,8 @@ def _giou(b1, b2):
         tf.zeros([tf.shape(illegal_area_indexes)[0], 1], tf.float64)
     ]
     iou = tf.dynamic_stitch(indices, data)
+    if mode == 'iou':
+        return iou
     bc_ymin = tf.minimum(b1_ymin, b2_ymin)
     bc_xmin = tf.minimum(b1_xmin, b2_xmin)
     bc_ymax = tf.maximum(b1_ymax, b2_ymax)
@@ -134,5 +149,6 @@ def _giou(b1, b2):
 
     enclose_area = tf.maximum(0, bc_ymax - bc_ymin) * tf.maximum(
         0, bc_xmax - bc_xmin)
-    giou = iou - (enclose_area - union_area) / enclose_area
+    giou = iou - tf.cast(enclose_area - union_area, tf.float64) / (
+        tf.cast(enclose_area, tf.float64) + 1e-8)
     return giou
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
index b53b70162a..5e06838bec 100644
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -1,3 +1,19 @@
+# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Tests for GIOU loss."""
+
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
@@ -9,11 +25,17 @@
 
 @test_utils.run_all_in_graph_and_eager_modes
 class GIOULossTest(tf.test.TestCase):
+    def test_config(self):
+        gl_obj = GIOULoss(
+            reduction=tf.keras.losses.Reduction.NONE, name='giou_loss')
+        self.assertEqual(gl_obj.name, 'giou_loss')
+        self.assertEqual(gl_obj.reduction, tf.keras.losses.Reduction.NONE)
+
     def test_giou_loss(self):
-        pred_data = [[0, 0, 1, 1], [0, 0, 2, 2]]
-        true_data = [[0, 0, 1, 1], [0, 0, 1, 2]]
+        pred_data = [[0, 0, 1, 1], [0, 0, 2, 2], [0, 0, 2, 2]]
+        true_data = [[0, 0, 1, 1], [0, 0, 1, 2], [2, 2, 4, 4]]
         loss = giou_loss(true_data, pred_data)
-        self.assertAllClose(loss, tf.constant([0, 0.5], tf.float64))
+        self.assertAllClose(loss, tf.constant([0, 0.5, 1.25], tf.float64))
 
 
 if __name__ == '__main__':

From ae1a0201b2ffb024abb4fe26462f3dbe9bdcfa84 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@outlook.com>
Date: Thu, 5 Sep 2019 10:53:34 +0800
Subject: [PATCH 03/22] Fix doc

---
 tensorflow_addons/losses/giou_loss.py | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index e4990ada48..12e2287603 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -33,13 +33,11 @@ class GIOULoss(tf.keras.losses.Loss):
     Usage:
 
     ```python
-    fl = tfa.losses.GIOU()
-    loss = fl(
-      [[0.97], [0.91], [0.03]],
-      [[1], [1], [0])
-    print('Loss: ', loss.numpy())  # Loss: [[0.03045921]
-                                            [0.09431068]
-                                            [0.31471074]
+    gl = tfa.losses.GIOU()
+    loss = gl(
+      [[0, 0, 1, 1], [0, 0, 2, 2], [0, 0, 2, 2]],
+      [[0, 0, 1, 1], [0, 0, 1, 2], [2, 2, 4, 4]])
+    print('Loss: ', loss.numpy())  # Loss: [0, 0.5, 1.25]
     ```
     Usage with tf.keras API:
 
@@ -49,6 +47,7 @@ class GIOULoss(tf.keras.losses.Loss):
     ```
 
     Args
+        reduction
       name: Op name
 
     Returns:

From 16064300e5461138121c6a82cfffdccfc264538a Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@outlook.com>
Date: Fri, 6 Sep 2019 09:29:14 +0800
Subject: [PATCH 04/22] Update Readme.md

---
 tensorflow_addons/losses/README.md | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
 mode change 100644 => 100755 tensorflow_addons/losses/README.md

diff --git a/tensorflow_addons/losses/README.md b/tensorflow_addons/losses/README.md
old mode 100644
new mode 100755
index ab54fef1bf..6aeb6c1f6c
--- a/tensorflow_addons/losses/README.md
+++ b/tensorflow_addons/losses/README.md
@@ -5,16 +5,17 @@
 |:---------- |:----------- |:------------- |
 | contrastive |  @WindQAQ | windqaq@gmail.com |
 | focal_loss | @SSaishruthi  | saishruthi.tn@gmail.com |
+| giou_loss | @who who who  | fsx950223@gmail.com |
 | lifted | @rahulunair | rahulunair@gmail.com  |
 | npairs | @WindQAQ | windqaq@gmail.com |
 | sparsemax_loss | @AndreasMadsen | amwwebdk+github@gmail.com |
-| triplet |  @rahulunair | rahulunair@gmail.com  |
 
 ## Components
 | Submodule | Loss  | Reference               |
 |:----------------------- |:---------------------|:--------------------------|
 | contrastive | ContrastiveLoss | http://yann.lecun.com/exdb/publis/pdf/hadsell-chopra-lecun-06.pdf |
 | focal_loss | SigmoidFocalCrossEntropy | https://arxiv.org/abs/1708.02002  |
+| giou_loss | GIOULoss | https://giou.stanford.edu/GIoU.pdf       |
 | lifted | LiftedStructLoss | https://arxiv.org/abs/1511.06452       |
 | npairs | NpairsLoss | http://www.nec-labs.com/uploads/images/Department-Images/MediaAnalytics/papers/nips16_npairmetriclearning.pdf |
 | npairs | NpairsMultilabelLoss | http://www.nec-labs.com/uploads/images/Department-Images/MediaAnalytics/papers/nips16_npairmetriclearning.pdf |

From eb7458dfd8b427408ced46a647a6fc5795c7a710 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@outlook.com>
Date: Mon, 9 Sep 2019 09:34:52 +0800
Subject: [PATCH 05/22] Format code

---
 tensorflow_addons/losses/giou_loss.py      | 8 ++++----
 tensorflow_addons/losses/giou_loss_test.py | 2 ++
 2 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index 12e2287603..add9045eca 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -26,9 +26,9 @@
 class GIOULoss(tf.keras.losses.Loss):
     """Implements the GIOU loss function.
 
-    GIOU loss was first introduced in the Generalized Intersection over Union paper
-    (https://giou.stanford.edu/GIoU.pdf). GIOU is a enhance for model which use IOU
-    in object detection.
+    GIOU loss was first introduced in the Generalized Intersection over Union
+    paper (https://giou.stanford.edu/GIoU.pdf). GIOU is a enhance for model
+    which use IOU in object detection.
 
     Usage:
 
@@ -73,7 +73,7 @@ def giou_loss(y_true, y_pred):
     Args
         y_true: true targets tensor.
         y_pred: predictions tensor.
-    
+
     Returns:
         GIOU loss float `Tensor`.
     """
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
index 5e06838bec..f5ae519f5e 100644
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -25,6 +25,8 @@
 
 @test_utils.run_all_in_graph_and_eager_modes
 class GIOULossTest(tf.test.TestCase):
+    """GIOU test class."""
+
     def test_config(self):
         gl_obj = GIOULoss(
             reduction=tf.keras.losses.Reduction.NONE, name='giou_loss')

From be8dcf09b6cf8f81eae71f129a55e014c5ce2e37 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Mon, 4 Nov 2019 04:07:28 +0000
Subject: [PATCH 06/22] refact calculate

---
 tensorflow_addons/losses/__init__.py       |   2 +-
 tensorflow_addons/losses/giou_loss.py      | 115 +++++++++------------
 tensorflow_addons/losses/giou_loss_test.py |  55 ++++++++--
 3 files changed, 97 insertions(+), 75 deletions(-)
 mode change 100644 => 100755 tensorflow_addons/losses/giou_loss.py
 mode change 100644 => 100755 tensorflow_addons/losses/giou_loss_test.py

diff --git a/tensorflow_addons/losses/__init__.py b/tensorflow_addons/losses/__init__.py
index efb00ec7a6..3db3aad827 100644
--- a/tensorflow_addons/losses/__init__.py
+++ b/tensorflow_addons/losses/__init__.py
@@ -24,4 +24,4 @@
 from tensorflow_addons.losses.npairs import npairs_loss, NpairsLoss, npairs_multilabel_loss, NpairsMultilabelLoss
 from tensorflow_addons.losses.sparsemax_loss import sparsemax_loss, SparsemaxLoss
 from tensorflow_addons.losses.triplet import triplet_semihard_loss, TripletSemiHardLoss
-from tensorflow_addons.losses.giou_loss import giou_loss, GIOULoss
+from tensorflow_addons.losses.giou_loss import giou_loss, GIoULoss
diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
old mode 100644
new mode 100755
index add9045eca..c79be5c8d5
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-"""Implements GIOU loss."""
+"""Implements GIoU loss."""
 
 from __future__ import absolute_import
 from __future__ import division
@@ -23,91 +23,83 @@
 
 
 @keras_utils.register_keras_custom_object
-class GIOULoss(tf.keras.losses.Loss):
-    """Implements the GIOU loss function.
+class GIoULoss(tf.keras.losses.Loss):
+    """Implements the GIoU loss function.
 
-    GIOU loss was first introduced in the Generalized Intersection over Union
-    paper (https://giou.stanford.edu/GIoU.pdf). GIOU is a enhance for model
+    GIoU loss was first introduced in the Generalized Intersection over Union
+    paper (https://giou.stanford.edu/GIoU.pdf). GIoU is a enhance for model
     which use IOU in object detection.
 
     Usage:
 
     ```python
-    gl = tfa.losses.GIOU()
-    loss = gl(
-      [[0, 0, 1, 1], [0, 0, 2, 2], [0, 0, 2, 2]],
-      [[0, 0, 1, 1], [0, 0, 1, 2], [2, 2, 4, 4]])
-    print('Loss: ', loss.numpy())  # Loss: [0, 0.5, 1.25]
+    gl = tfa.losses.GIoU()
+    boxes1 = tf.constant([[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]])
+    boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]])
+    loss = gl(boxes1,boxes2)
+    print('Loss: ', loss.numpy())  # Loss: [1.07500000298023224, 1.9333333373069763]
     ```
     Usage with tf.keras API:
 
     ```python
     model = tf.keras.Model(inputs, outputs)
-    model.compile('sgd', loss=tf.keras.losses.GIOULoss())
+    model.compile('sgd', loss=tf.keras.losses.GIoULoss())
     ```
 
     Args
-        reduction
-      name: Op name
+      mode: one of ['giou', 'iou'], decided to calculate giou loss or iou loss.
 
     Returns:
-      GIOU loss float `Tensor`.
+      GIoU loss double `Tensor`.
     """
 
     def __init__(self,
-                 reduction=tf.keras.losses.Reduction.NONE,
+                 mode='giou',
+                 reduction=tf.keras.losses.Reduction.AUTO,
                  name='giou_loss'):
-        super(GIOULoss, self).__init__(name=name, reduction=reduction)
-
-    def call(self, y_true, y_pred):
-        return giou_loss(y_true, y_pred)
+        super(GIoULoss, self).__init__(name=name, reduction=reduction)
+        self.mode = mode
 
     def get_config(self):
-        base_config = super(GIOULoss, self).get_config()
+        base_config = super(GIoULoss, self).get_config()
+        base_config['mode'] = self.mode
         return base_config
 
+    def call(self, y_true, y_pred):
+        return giou_loss(y_true, y_pred, mode=self.mode)
+
 
 @keras_utils.register_keras_custom_object
-def giou_loss(y_true, y_pred):
+@tf.function
+def giou_loss(y_true, y_pred, mode='giou'):
     """
     Args
         y_true: true targets tensor.
         y_pred: predictions tensor.
 
     Returns:
-        GIOU loss float `Tensor`.
+        GIoU loss double `Tensor`.
     """
 
     y_pred = tf.convert_to_tensor(y_pred)
     y_true = tf.cast(y_true, y_pred.dtype)
-    giou = giou_calculate(y_pred, y_true)
+    giou = do_giou_calculate(y_pred, y_true, mode)
 
     # compute the final loss and return
     return 1 - giou
 
 
-def giou_calculate(b1, b2):
+def do_giou_calculate(b1, b2, mode='giou'):
     """
     Args
         b1: bbox.
-        b1: the other bbox.
+        b2: the other bbox.
+        mode: one of ['giou', 'iou']
 
     Returns:
-        iou float `Tensor`, between [-1, 1].
-    """
-    return do_iou_calculate(b1, b2, mode='giou')
-
-
-def do_iou_calculate(b1, b2, mode='iou'):
-    """
-    Args
-        b1: bbox.
-        b1: the other bbox.
-        mode: one of ['iou', 'giou']
-
-    Returns:
-        iou float `Tensor`.
+        GIoU loss double `Tensor`.
     """
+    zero = tf.convert_to_tensor(0., b1.dtype)
     b1_ymin = tf.minimum(b1[:, 0], b1[:, 2])
     b1_xmin = tf.minimum(b1[:, 1], b1[:, 3])
     b1_ymax = tf.maximum(b1[:, 0], b1[:, 2])
@@ -116,38 +108,33 @@ def do_iou_calculate(b1, b2, mode='iou'):
     b2_xmin = tf.minimum(b2[:, 1], b2[:, 3])
     b2_ymax = tf.maximum(b2[:, 0], b2[:, 2])
     b2_xmax = tf.maximum(b2[:, 1], b2[:, 3])
-    b1_area = (b1_ymax - b1_ymin) * (b1_xmax - b1_xmin)
-    b2_area = (b2_ymax - b2_ymin) * (b2_xmax - b1_xmin)
-    illegal_area_indexes = tf.cast(
-        tf.where(tf.logical_or(b1_area < 0, b2_area < 0)), tf.int32)
-    valid_area_indexes = tf.cast(
-        tf.where(tf.logical_and(b1_area >= 0, b2_area >= 0)), tf.int32)
+    b1_width = tf.maximum(zero, b1_xmax - b1_xmin)
+    b1_height = tf.maximum(zero, b1_ymax - b1_ymin)
+    b2_width = tf.maximum(zero, b2_xmax - b2_xmin)
+    b2_height = tf.maximum(zero, b2_ymax - b2_ymin)
+    b1_area = b1_width * b1_height
+    b2_area = b2_width * b2_height
 
     intersect_ymin = tf.maximum(b1_ymin, b2_ymin)
     intersect_xmin = tf.maximum(b1_xmin, b2_xmin)
     intersect_ymax = tf.minimum(b1_ymax, b2_ymax)
     intersect_xmax = tf.minimum(b1_xmax, b2_xmax)
-    intersect_area = tf.maximum(0,
-                                intersect_ymax - intersect_ymin) * tf.maximum(
-                                    0, intersect_xmax - intersect_xmin)
+    intersect_width = tf.maximum(zero, intersect_xmax - intersect_xmin)
+    intersect_height = tf.maximum(zero, intersect_ymax - intersect_ymin)
+    intersect_area = intersect_width * intersect_height
 
     union_area = b1_area + b2_area - intersect_area
-    iou = intersect_area / union_area
-    indices = [valid_area_indexes, illegal_area_indexes]
-    data = [
-        tf.gather(iou, valid_area_indexes),
-        tf.zeros([tf.shape(illegal_area_indexes)[0], 1], tf.float64)
-    ]
-    iou = tf.dynamic_stitch(indices, data)
+    iou = tf.math.divide_no_nan(intersect_area, union_area)
     if mode == 'iou':
         return iou
-    bc_ymin = tf.minimum(b1_ymin, b2_ymin)
-    bc_xmin = tf.minimum(b1_xmin, b2_xmin)
-    bc_ymax = tf.maximum(b1_ymax, b2_ymax)
-    bc_xmax = tf.maximum(b1_xmax, b2_xmax)
-
-    enclose_area = tf.maximum(0, bc_ymax - bc_ymin) * tf.maximum(
-        0, bc_xmax - bc_xmin)
-    giou = iou - tf.cast(enclose_area - union_area, tf.float64) / (
-        tf.cast(enclose_area, tf.float64) + 1e-8)
+
+    enclose_ymin = tf.minimum(b1_ymin, b2_ymin)
+    enclose_xmin = tf.minimum(b1_xmin, b2_xmin)
+    enclose_ymax = tf.maximum(b1_ymax, b2_ymax)
+    enclose_xmax = tf.maximum(b1_xmax, b2_xmax)
+    enclose_width = tf.maximum(zero, enclose_xmax - enclose_xmin)
+    enclose_height = tf.maximum(zero, enclose_ymax - enclose_ymin)
+    enclose_area = enclose_width * enclose_height
+    giou = iou - tf.math.divide_no_nan(
+        (enclose_area - union_area), enclose_area)
     return giou
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
old mode 100644
new mode 100755
index f5ae519f5e..b95fb769bd
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -12,32 +12,67 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-"""Tests for GIOU loss."""
+"""Tests for GIoU loss."""
 
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
+from absl.testing import parameterized
+
+import numpy as np
 import tensorflow as tf
 from tensorflow_addons.utils import test_utils
-from tensorflow_addons.losses import giou_loss, GIOULoss
+from tensorflow_addons.losses import giou_loss, GIoULoss
 
 
 @test_utils.run_all_in_graph_and_eager_modes
-class GIOULossTest(tf.test.TestCase):
-    """GIOU test class."""
+class GIoULossTest(tf.test.TestCase, parameterized.TestCase):
+    """GIoU test class."""
 
     def test_config(self):
-        gl_obj = GIOULoss(
+        gl_obj = GIoULoss(
             reduction=tf.keras.losses.Reduction.NONE, name='giou_loss')
         self.assertEqual(gl_obj.name, 'giou_loss')
         self.assertEqual(gl_obj.reduction, tf.keras.losses.Reduction.NONE)
 
-    def test_giou_loss(self):
-        pred_data = [[0, 0, 1, 1], [0, 0, 2, 2], [0, 0, 2, 2]]
-        true_data = [[0, 0, 1, 1], [0, 0, 1, 2], [2, 2, 4, 4]]
-        loss = giou_loss(true_data, pred_data)
-        self.assertAllClose(loss, tf.constant([0, 0.5, 1.25], tf.float64))
+    @parameterized.named_parameters(("float16", np.float16),
+                                    ("float32", np.float32),
+                                    ("float64", np.float64))
+    def test_iou(self, dtype):
+        boxes1 = tf.constant([[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
+                             dtype=dtype)
+        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]],
+                             dtype=dtype)
+        expected_result = tf.constant([14.0 / 16.0, 1.], dtype=dtype)
+        loss = giou_loss(boxes1, boxes2, mode='iou')
+        self.assertAllCloseAccordingToType(loss, expected_result)
+
+    @parameterized.named_parameters(("float16", np.float16),
+                                    ("float32", np.float32),
+                                    ("float64", np.float64))
+    def test_giou_loss(self, dtype):
+        boxes1 = tf.constant([[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
+                             dtype=dtype)
+        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]],
+                             dtype=dtype)
+        expected_result = tf.constant(
+            [1.07500000298023224, 1.9333333373069763], dtype=dtype)
+        loss = giou_loss(boxes1, boxes2)
+        self.assertAllCloseAccordingToType(loss, expected_result)
+
+    @parameterized.named_parameters(("float16", np.float16),
+                                    ("float32", np.float32),
+                                    ("float64", np.float64))
+    def test_giou_loss_swapped(self, dtype):
+        boxes1 = tf.constant([[7.0, 3.0, 4.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
+                             dtype=dtype)
+        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]],
+                             dtype=dtype)
+        expected_result = tf.constant(
+            [1.07500000298023224, 1.9333333373069763], dtype=dtype)
+        loss = giou_loss(boxes1, boxes2)
+        self.assertAllCloseAccordingToType(loss, expected_result)
 
 
 if __name__ == '__main__':

From 7f72cab314062702d62933dfc15c78678d7c05ba Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Mon, 4 Nov 2019 07:25:31 +0000
Subject: [PATCH 07/22] fix document

---
 tensorflow_addons/losses/README.md         |  2 +-
 tensorflow_addons/losses/giou_loss.py      | 22 ++++++++++++----------
 tensorflow_addons/losses/giou_loss_test.py |  0
 3 files changed, 13 insertions(+), 11 deletions(-)
 mode change 100755 => 100644 tensorflow_addons/losses/giou_loss.py
 mode change 100755 => 100644 tensorflow_addons/losses/giou_loss_test.py

diff --git a/tensorflow_addons/losses/README.md b/tensorflow_addons/losses/README.md
index 6aeb6c1f6c..1d846e1132 100755
--- a/tensorflow_addons/losses/README.md
+++ b/tensorflow_addons/losses/README.md
@@ -15,7 +15,7 @@
 |:----------------------- |:---------------------|:--------------------------|
 | contrastive | ContrastiveLoss | http://yann.lecun.com/exdb/publis/pdf/hadsell-chopra-lecun-06.pdf |
 | focal_loss | SigmoidFocalCrossEntropy | https://arxiv.org/abs/1708.02002  |
-| giou_loss | GIOULoss | https://giou.stanford.edu/GIoU.pdf       |
+| giou_loss | GIoULoss | https://giou.stanford.edu/GIoU.pdf       |
 | lifted | LiftedStructLoss | https://arxiv.org/abs/1511.06452       |
 | npairs | NpairsLoss | http://www.nec-labs.com/uploads/images/Department-Images/MediaAnalytics/papers/nips16_npairmetriclearning.pdf |
 | npairs | NpairsMultilabelLoss | http://www.nec-labs.com/uploads/images/Department-Images/MediaAnalytics/papers/nips16_npairmetriclearning.pdf |
diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
old mode 100755
new mode 100644
index c79be5c8d5..cf2cd3acb8
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -46,11 +46,11 @@ class GIoULoss(tf.keras.losses.Loss):
     model.compile('sgd', loss=tf.keras.losses.GIoULoss())
     ```
 
-    Args
+    Args:
       mode: one of ['giou', 'iou'], decided to calculate giou loss or iou loss.
 
     Returns:
-      GIoU loss double `Tensor`.
+      GIoU loss float `Tensor`.
     """
 
     def __init__(self,
@@ -73,14 +73,16 @@ def call(self, y_true, y_pred):
 @tf.function
 def giou_loss(y_true, y_pred, mode='giou'):
     """
-    Args
+    Args:
         y_true: true targets tensor.
         y_pred: predictions tensor.
+        mode: one of ['giou', 'iou'], decided to calculate giou loss or iou loss.
 
     Returns:
-        GIoU loss double `Tensor`.
+        GIoU loss float `Tensor`.
     """
-
+    if mode not in ['giou', 'iou']:
+        raise ValueError("Value of mode should be 'iou' or 'giou'")
     y_pred = tf.convert_to_tensor(y_pred)
     y_true = tf.cast(y_true, y_pred.dtype)
     giou = do_giou_calculate(y_pred, y_true, mode)
@@ -91,13 +93,13 @@ def giou_loss(y_true, y_pred, mode='giou'):
 
 def do_giou_calculate(b1, b2, mode='giou'):
     """
-    Args
-        b1: bbox.
-        b2: the other bbox.
-        mode: one of ['giou', 'iou']
+    Args:
+        b1: bounding box.
+        b2: the other bounding box.
+        mode: one of ['giou', 'iou'], decided to calculate giou loss or iou loss.
 
     Returns:
-        GIoU loss double `Tensor`.
+        GIoU loss float `Tensor`.
     """
     zero = tf.convert_to_tensor(0., b1.dtype)
     b1_ymin = tf.minimum(b1[:, 0], b1[:, 2])
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
old mode 100755
new mode 100644

From 099298c4b8901c5ddf195d2c7e84ac630e6be795 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Tue, 5 Nov 2019 01:58:43 +0000
Subject: [PATCH 08/22] fix readme

---
 tensorflow_addons/losses/README.md | 1 +
 1 file changed, 1 insertion(+)

diff --git a/tensorflow_addons/losses/README.md b/tensorflow_addons/losses/README.md
index 1d846e1132..ec570a51a8 100755
--- a/tensorflow_addons/losses/README.md
+++ b/tensorflow_addons/losses/README.md
@@ -9,6 +9,7 @@
 | lifted | @rahulunair | rahulunair@gmail.com  |
 | npairs | @WindQAQ | windqaq@gmail.com |
 | sparsemax_loss | @AndreasMadsen | amwwebdk+github@gmail.com |
+| triplet |  @rahulunair | rahulunair@gmail.com  |
 
 ## Components
 | Submodule | Loss  | Reference               |

From 825c255fe126fbc71a83706e1f8fbf392d122d7f Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Tue, 5 Nov 2019 02:03:30 +0000
Subject: [PATCH 09/22] fix docs

---
 tensorflow_addons/losses/README.md    | 0
 tensorflow_addons/losses/giou_loss.py | 6 +++---
 2 files changed, 3 insertions(+), 3 deletions(-)
 mode change 100755 => 100644 tensorflow_addons/losses/README.md

diff --git a/tensorflow_addons/losses/README.md b/tensorflow_addons/losses/README.md
old mode 100755
new mode 100644
diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index cf2cd3acb8..e242ec2f81 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -26,9 +26,9 @@
 class GIoULoss(tf.keras.losses.Loss):
     """Implements the GIoU loss function.
 
-    GIoU loss was first introduced in the Generalized Intersection over Union
-    paper (https://giou.stanford.edu/GIoU.pdf). GIoU is a enhance for model
-    which use IOU in object detection.
+    GIoU loss was first introduced in the
+    [Generalized Intersection over Union paper](https://giou.stanford.edu/GIoU.pdf).
+    GIoU is a enhance for model which use IOU in object detection.
 
     Usage:
 

From 9beec61b9016a28463b71a03ce810b08d42294c6 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Tue, 5 Nov 2019 02:14:25 +0000
Subject: [PATCH 10/22] Change to official api

---
 tensorflow_addons/losses/giou_loss.py | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index e242ec2f81..9625822a16 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -19,10 +19,9 @@
 from __future__ import print_function
 
 import tensorflow as tf
-from tensorflow_addons.utils import keras_utils
 
 
-@keras_utils.register_keras_custom_object
+@tf.keras.utils.register_keras_serializable(package='Addons')
 class GIoULoss(tf.keras.losses.Loss):
     """Implements the GIoU loss function.
 
@@ -69,7 +68,7 @@ def call(self, y_true, y_pred):
         return giou_loss(y_true, y_pred, mode=self.mode)
 
 
-@keras_utils.register_keras_custom_object
+@tf.keras.utils.register_keras_serializable(package='Addons')
 @tf.function
 def giou_loss(y_true, y_pred, mode='giou'):
     """

From 72fc21a8986380acbc6195ea20f51c80c387a7c1 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Tue, 5 Nov 2019 04:45:54 +0000
Subject: [PATCH 11/22] format code

---
 tensorflow_addons/losses/giou_loss.py | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index 9625822a16..4cc3ae4f00 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -75,7 +75,8 @@ def giou_loss(y_true, y_pred, mode='giou'):
     Args:
         y_true: true targets tensor.
         y_pred: predictions tensor.
-        mode: one of ['giou', 'iou'], decided to calculate giou loss or iou loss.
+        mode: one of ['giou', 'iou'],
+        decided to calculate giou loss or iou loss.
 
     Returns:
         GIoU loss float `Tensor`.
@@ -95,7 +96,8 @@ def do_giou_calculate(b1, b2, mode='giou'):
     Args:
         b1: bounding box.
         b2: the other bounding box.
-        mode: one of ['giou', 'iou'], decided to calculate giou loss or iou loss.
+        mode: one of ['giou', 'iou'],
+        decided to calculate giou loss or iou loss.
 
     Returns:
         GIoU loss float `Tensor`.

From b35dff90d74df79d5c9ecb36516e3feda8e0cf41 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Wed, 6 Nov 2019 02:04:02 +0000
Subject: [PATCH 12/22] enhance robust

---
 tensorflow_addons/losses/giou_loss.py      | 16 ++++++++--------
 tensorflow_addons/losses/giou_loss_test.py | 18 ++++++++++++++++++
 2 files changed, 26 insertions(+), 8 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index 4cc3ae4f00..f42069e06e 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -103,14 +103,14 @@ def do_giou_calculate(b1, b2, mode='giou'):
         GIoU loss float `Tensor`.
     """
     zero = tf.convert_to_tensor(0., b1.dtype)
-    b1_ymin = tf.minimum(b1[:, 0], b1[:, 2])
-    b1_xmin = tf.minimum(b1[:, 1], b1[:, 3])
-    b1_ymax = tf.maximum(b1[:, 0], b1[:, 2])
-    b1_xmax = tf.maximum(b1[:, 1], b1[:, 3])
-    b2_ymin = tf.minimum(b2[:, 0], b2[:, 2])
-    b2_xmin = tf.minimum(b2[:, 1], b2[:, 3])
-    b2_ymax = tf.maximum(b2[:, 0], b2[:, 2])
-    b2_xmax = tf.maximum(b2[:, 1], b2[:, 3])
+    b1_ymin = tf.minimum(b1[..., 0], b1[..., 2])
+    b1_xmin = tf.minimum(b1[..., 1], b1[..., 3])
+    b1_ymax = tf.maximum(b1[..., 0], b1[..., 2])
+    b1_xmax = tf.maximum(b1[..., 1], b1[..., 3])
+    b2_ymin = tf.minimum(b2[..., 0], b2[..., 2])
+    b2_xmin = tf.minimum(b2[..., 1], b2[..., 3])
+    b2_ymax = tf.maximum(b2[..., 0], b2[..., 2])
+    b2_xmax = tf.maximum(b2[..., 1], b2[..., 3])
     b1_width = tf.maximum(zero, b1_xmax - b1_xmin)
     b1_height = tf.maximum(zero, b1_ymax - b1_ymin)
     b2_width = tf.maximum(zero, b2_xmax - b2_xmin)
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
index b95fb769bd..5909713822 100644
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -74,6 +74,24 @@ def test_giou_loss_swapped(self, dtype):
         loss = giou_loss(boxes1, boxes2)
         self.assertAllCloseAccordingToType(loss, expected_result)
 
+    @parameterized.named_parameters(("float16", np.float16),
+                                    ("float32", np.float32),
+                                    ("float64", np.float64))
+    def test_giou_loss_with_dims(self, dtype):
+        boxes1 = tf.constant([[[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
+                              [[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]]],
+                             dtype=dtype)
+        boxes2 = tf.constant(
+            [[[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]],
+             [[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]]],
+            dtype=dtype)
+        expected_result = tf.constant(
+            [[1.07500000298023224, 1.9333333373069763],
+             [1.07500000298023224, 1.9333333373069763]],
+            dtype=dtype)
+        loss = giou_loss(boxes1, boxes2)
+        self.assertAllCloseAccordingToType(loss, expected_result)
+
 
 if __name__ == '__main__':
     tf.test.main()

From 310850df438a2813c0f13b9417df538cbfb68f3e Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Sun, 10 Nov 2019 04:19:57 +0000
Subject: [PATCH 13/22] add box format

---
 tensorflow_addons/losses/giou_loss.py      | 22 ++++++++++------------
 tensorflow_addons/losses/giou_loss_test.py | 13 -------------
 2 files changed, 10 insertions(+), 25 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index f42069e06e..a584c56871 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -73,8 +73,10 @@ def call(self, y_true, y_pred):
 def giou_loss(y_true, y_pred, mode='giou'):
     """
     Args:
-        y_true: true targets tensor.
-        y_pred: predictions tensor.
+        y_true: true targets tensor. The coordinates of the each bounding box in
+        boxes are encoded as [y_min, x_min, y_max, x_max].
+        y_pred: predictions tensor. The coordinates of the each bounding box in
+        boxes are encoded as [y_min, x_min, y_max, x_max].
         mode: one of ['giou', 'iou'],
         decided to calculate giou loss or iou loss.
 
@@ -94,8 +96,10 @@ def giou_loss(y_true, y_pred, mode='giou'):
 def do_giou_calculate(b1, b2, mode='giou'):
     """
     Args:
-        b1: bounding box.
-        b2: the other bounding box.
+        b1: bounding box. The coordinates of the each bounding box in boxes are
+        encoded as [y_min, x_min, y_max, x_max].
+        b2: the other bounding box. The coordinates of the each bounding box 
+        in boxes are encoded as [y_min, x_min, y_max, x_max].
         mode: one of ['giou', 'iou'],
         decided to calculate giou loss or iou loss.
 
@@ -103,14 +107,8 @@ def do_giou_calculate(b1, b2, mode='giou'):
         GIoU loss float `Tensor`.
     """
     zero = tf.convert_to_tensor(0., b1.dtype)
-    b1_ymin = tf.minimum(b1[..., 0], b1[..., 2])
-    b1_xmin = tf.minimum(b1[..., 1], b1[..., 3])
-    b1_ymax = tf.maximum(b1[..., 0], b1[..., 2])
-    b1_xmax = tf.maximum(b1[..., 1], b1[..., 3])
-    b2_ymin = tf.minimum(b2[..., 0], b2[..., 2])
-    b2_xmin = tf.minimum(b2[..., 1], b2[..., 3])
-    b2_ymax = tf.maximum(b2[..., 0], b2[..., 2])
-    b2_xmax = tf.maximum(b2[..., 1], b2[..., 3])
+    b1_ymin, b1_xmin, b1_ymax, b1_xmax = tf.unstack(b1, axis=-1)
+    b2_ymin, b2_xmin, b2_ymax, b2_xmax = tf.unstack(b2, axis=-1)
     b1_width = tf.maximum(zero, b1_xmax - b1_xmin)
     b1_height = tf.maximum(zero, b1_ymax - b1_ymin)
     b2_width = tf.maximum(zero, b2_xmax - b2_xmin)
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
index 5909713822..c61345cb32 100644
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -61,19 +61,6 @@ def test_giou_loss(self, dtype):
         loss = giou_loss(boxes1, boxes2)
         self.assertAllCloseAccordingToType(loss, expected_result)
 
-    @parameterized.named_parameters(("float16", np.float16),
-                                    ("float32", np.float32),
-                                    ("float64", np.float64))
-    def test_giou_loss_swapped(self, dtype):
-        boxes1 = tf.constant([[7.0, 3.0, 4.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
-                             dtype=dtype)
-        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]],
-                             dtype=dtype)
-        expected_result = tf.constant(
-            [1.07500000298023224, 1.9333333373069763], dtype=dtype)
-        loss = giou_loss(boxes1, boxes2)
-        self.assertAllCloseAccordingToType(loss, expected_result)
-
     @parameterized.named_parameters(("float16", np.float16),
                                     ("float32", np.float32),
                                     ("float64", np.float64))

From f85f692080843ac34fc53475d362526e18e0adab Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Tue, 12 Nov 2019 07:17:19 +0000
Subject: [PATCH 14/22] add keras test

---
 tensorflow_addons/losses/__init__.py       |  2 +-
 tensorflow_addons/losses/giou_loss.py      |  9 ++++-----
 tensorflow_addons/losses/giou_loss_test.py | 21 ++++++++++-----------
 3 files changed, 15 insertions(+), 17 deletions(-)

diff --git a/tensorflow_addons/losses/__init__.py b/tensorflow_addons/losses/__init__.py
index 3db3aad827..7aa292527f 100644
--- a/tensorflow_addons/losses/__init__.py
+++ b/tensorflow_addons/losses/__init__.py
@@ -20,8 +20,8 @@
 
 from tensorflow_addons.losses.contrastive import contrastive_loss, ContrastiveLoss
 from tensorflow_addons.losses.focal_loss import sigmoid_focal_crossentropy, SigmoidFocalCrossEntropy
+from tensorflow_addons.losses.giou_loss import giou_loss, GIoULoss
 from tensorflow_addons.losses.lifted import lifted_struct_loss, LiftedStructLoss
 from tensorflow_addons.losses.npairs import npairs_loss, NpairsLoss, npairs_multilabel_loss, NpairsMultilabelLoss
 from tensorflow_addons.losses.sparsemax_loss import sparsemax_loss, SparsemaxLoss
 from tensorflow_addons.losses.triplet import triplet_semihard_loss, TripletSemiHardLoss
-from tensorflow_addons.losses.giou_loss import giou_loss, GIoULoss
diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index a584c56871..276156446c 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -47,15 +47,14 @@ class GIoULoss(tf.keras.losses.Loss):
 
     Args:
       mode: one of ['giou', 'iou'], decided to calculate giou loss or iou loss.
-
-    Returns:
-      GIoU loss float `Tensor`.
     """
 
     def __init__(self,
                  mode='giou',
                  reduction=tf.keras.losses.Reduction.AUTO,
                  name='giou_loss'):
+        if mode not in ['giou', 'iou']:
+            raise ValueError("Value of mode should be 'iou' or 'giou'")
         super(GIoULoss, self).__init__(name=name, reduction=reduction)
         self.mode = mode
 
@@ -107,8 +106,8 @@ def do_giou_calculate(b1, b2, mode='giou'):
         GIoU loss float `Tensor`.
     """
     zero = tf.convert_to_tensor(0., b1.dtype)
-    b1_ymin, b1_xmin, b1_ymax, b1_xmax = tf.unstack(b1, axis=-1)
-    b2_ymin, b2_xmin, b2_ymax, b2_xmax = tf.unstack(b2, axis=-1)
+    b1_ymin, b1_xmin, b1_ymax, b1_xmax = tf.unstack(b1, 4, axis=-1)
+    b2_ymin, b2_xmin, b2_ymax, b2_xmax = tf.unstack(b2, 4, axis=-1)
     b1_width = tf.maximum(zero, b1_xmax - b1_xmin)
     b1_height = tf.maximum(zero, b1_ymax - b1_ymin)
     b2_width = tf.maximum(zero, b2_xmax - b2_xmin)
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
index c61345cb32..e52c1c2c08 100644
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -64,19 +64,18 @@ def test_giou_loss(self, dtype):
     @parameterized.named_parameters(("float16", np.float16),
                                     ("float32", np.float32),
                                     ("float64", np.float64))
-    def test_giou_loss_with_dims(self, dtype):
-        boxes1 = tf.constant([[[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
-                              [[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]]],
+    def test_keras_model(self, dtype):
+        boxes1 = tf.constant([[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
+                             dtype=dtype)
+        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]],
                              dtype=dtype)
-        boxes2 = tf.constant(
-            [[[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]],
-             [[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]]],
-            dtype=dtype)
         expected_result = tf.constant(
-            [[1.07500000298023224, 1.9333333373069763],
-             [1.07500000298023224, 1.9333333373069763]],
-            dtype=dtype)
-        loss = giou_loss(boxes1, boxes2)
+            [1.07500000298023224, 1.9333333373069763], dtype=dtype)
+        model = tf.keras.Sequential()
+        model.compile(
+            optimizer='adam',
+            loss=GIoULoss(reduction=tf.keras.losses.Reduction.NONE))
+        loss = model.evaluate(boxes1, boxes2, batch_size=2, steps=1)
         self.assertAllCloseAccordingToType(loss, expected_result)
 
 

From ec11713a8bd5365932e565509aca5c079a197284 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Wed, 13 Nov 2019 02:57:11 +0000
Subject: [PATCH 15/22] add one bbox test

---
 tensorflow_addons/losses/giou_loss.py      |  1 +
 tensorflow_addons/losses/giou_loss_test.py | 10 ++++++++++
 2 files changed, 11 insertions(+)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index 276156446c..6d413cad33 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -82,6 +82,7 @@ def giou_loss(y_true, y_pred, mode='giou'):
     Returns:
         GIoU loss float `Tensor`.
     """
+    # if y_true or y_pred:
     if mode not in ['giou', 'iou']:
         raise ValueError("Value of mode should be 'iou' or 'giou'")
     y_pred = tf.convert_to_tensor(y_pred)
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
index e52c1c2c08..e09e28edbe 100644
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -61,6 +61,16 @@ def test_giou_loss(self, dtype):
         loss = giou_loss(boxes1, boxes2)
         self.assertAllCloseAccordingToType(loss, expected_result)
 
+    @parameterized.named_parameters(("float16", np.float16),
+                                    ("float32", np.float32),
+                                    ("float64", np.float64))
+    def test_one_bbox(self, dtype):
+        boxes1 = tf.constant([4.0, 3.0, 7.0, 5.0], dtype=dtype)
+        boxes2 = tf.constant([3.0, 4.0, 6.0, 8.0], dtype=dtype)
+        expected_result = tf.constant(1.07500000298023224, dtype=dtype)
+        loss = giou_loss(boxes1, boxes2)
+        self.assertAllCloseAccordingToType(loss, expected_result)
+
     @parameterized.named_parameters(("float16", np.float16),
                                     ("float32", np.float32),
                                     ("float64", np.float64))

From 8d4b3dd0b19d0374f238a0bd3996788da06b4eea Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Wed, 13 Nov 2019 05:28:26 +0000
Subject: [PATCH 16/22] add different shapes test case

---
 tensorflow_addons/losses/giou_loss_test.py | 26 ++++++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
index e09e28edbe..b804e97549 100644
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -61,6 +61,32 @@ def test_giou_loss(self, dtype):
         loss = giou_loss(boxes1, boxes2)
         self.assertAllCloseAccordingToType(loss, expected_result)
 
+    @parameterized.named_parameters(("float16", np.float16),
+                                ("float32", np.float32),
+                                ("float64", np.float64))
+    def test_different_shapes(self, dtype):
+        boxes1 = tf.constant([[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
+                             dtype=dtype)
+        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0]],
+                             dtype=dtype)
+        tf.expand_dims(boxes1,-2)
+        tf.expand_dims(boxes2,0)
+        expected_result = tf.constant(
+            [1.07500000298023224, 1.366071], dtype=dtype)
+        loss = giou_loss(boxes1, boxes2)
+        self.assertAllCloseAccordingToType(loss, expected_result)
+
+        boxes1 = tf.constant([[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
+                             dtype=dtype)
+        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0]],
+                             dtype=dtype)
+        tf.expand_dims(boxes1,0)
+        tf.expand_dims(boxes2,-2)
+        expected_result = tf.constant(
+            [1.07500000298023224, 1.366071], dtype=dtype)
+        loss = giou_loss(boxes1, boxes2)
+        self.assertAllCloseAccordingToType(loss, expected_result)
+
     @parameterized.named_parameters(("float16", np.float16),
                                     ("float32", np.float32),
                                     ("float64", np.float64))

From 55f4d0b2a1dd0d4375c886f62e110ab68718048c Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Wed, 13 Nov 2019 05:34:52 +0000
Subject: [PATCH 17/22] format code

---
 tensorflow_addons/losses/giou_loss.py      | 10 ++++-----
 tensorflow_addons/losses/giou_loss_test.py | 26 ++++++++++------------
 2 files changed, 17 insertions(+), 19 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index 6d413cad33..0225d93f09 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -72,10 +72,10 @@ def call(self, y_true, y_pred):
 def giou_loss(y_true, y_pred, mode='giou'):
     """
     Args:
-        y_true: true targets tensor. The coordinates of the each bounding box in
-        boxes are encoded as [y_min, x_min, y_max, x_max].
-        y_pred: predictions tensor. The coordinates of the each bounding box in
-        boxes are encoded as [y_min, x_min, y_max, x_max].
+        y_true: true targets tensor. The coordinates of the each bounding
+        box in boxes are encoded as [y_min, x_min, y_max, x_max].
+        y_pred: predictions tensor. The coordinates of the each bounding
+        box in boxes are encoded as [y_min, x_min, y_max, x_max].
         mode: one of ['giou', 'iou'],
         decided to calculate giou loss or iou loss.
 
@@ -98,7 +98,7 @@ def do_giou_calculate(b1, b2, mode='giou'):
     Args:
         b1: bounding box. The coordinates of the each bounding box in boxes are
         encoded as [y_min, x_min, y_max, x_max].
-        b2: the other bounding box. The coordinates of the each bounding box 
+        b2: the other bounding box. The coordinates of the each bounding box
         in boxes are encoded as [y_min, x_min, y_max, x_max].
         mode: one of ['giou', 'iou'],
         decided to calculate giou loss or iou loss.
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
index b804e97549..e950079fe5 100644
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -62,28 +62,26 @@ def test_giou_loss(self, dtype):
         self.assertAllCloseAccordingToType(loss, expected_result)
 
     @parameterized.named_parameters(("float16", np.float16),
-                                ("float32", np.float32),
-                                ("float64", np.float64))
+                                    ("float32", np.float32),
+                                    ("float64", np.float64))
     def test_different_shapes(self, dtype):
         boxes1 = tf.constant([[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
                              dtype=dtype)
-        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0]],
-                             dtype=dtype)
-        tf.expand_dims(boxes1,-2)
-        tf.expand_dims(boxes2,0)
-        expected_result = tf.constant(
-            [1.07500000298023224, 1.366071], dtype=dtype)
+        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0]], dtype=dtype)
+        tf.expand_dims(boxes1, -2)
+        tf.expand_dims(boxes2, 0)
+        expected_result = tf.constant([1.07500000298023224, 1.366071],
+                                      dtype=dtype)
         loss = giou_loss(boxes1, boxes2)
         self.assertAllCloseAccordingToType(loss, expected_result)
 
         boxes1 = tf.constant([[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]],
                              dtype=dtype)
-        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0]],
-                             dtype=dtype)
-        tf.expand_dims(boxes1,0)
-        tf.expand_dims(boxes2,-2)
-        expected_result = tf.constant(
-            [1.07500000298023224, 1.366071], dtype=dtype)
+        boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0]], dtype=dtype)
+        tf.expand_dims(boxes1, 0)
+        tf.expand_dims(boxes2, -2)
+        expected_result = tf.constant([1.07500000298023224, 1.366071],
+                                      dtype=dtype)
         loss = giou_loss(boxes1, boxes2)
         self.assertAllCloseAccordingToType(loss, expected_result)
 

From f830058b94431f70f9d430386dc2877485ea31d1 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Sun, 17 Nov 2019 06:10:49 +0000
Subject: [PATCH 18/22] fix docs

---
 tensorflow_addons/losses/giou_loss.py | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index 0225d93f09..0de33ea88c 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -26,8 +26,10 @@ class GIoULoss(tf.keras.losses.Loss):
     """Implements the GIoU loss function.
 
     GIoU loss was first introduced in the
-    [Generalized Intersection over Union paper](https://giou.stanford.edu/GIoU.pdf).
-    GIoU is a enhance for model which use IOU in object detection.
+    [Generalized Intersection over Union:
+    A Metric and A Loss for Bounding Box Regression]
+    (https://giou.stanford.edu/GIoU.pdf).
+    GIoU is a enhance for model which use IoU in object detection.
 
     Usage:
 
@@ -35,18 +37,18 @@ class GIoULoss(tf.keras.losses.Loss):
     gl = tfa.losses.GIoU()
     boxes1 = tf.constant([[4.0, 3.0, 7.0, 5.0], [5.0, 6.0, 10.0, 7.0]])
     boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]])
-    loss = gl(boxes1,boxes2)
+    loss = gl(boxes1, boxes2)
     print('Loss: ', loss.numpy())  # Loss: [1.07500000298023224, 1.9333333373069763]
     ```
     Usage with tf.keras API:
 
     ```python
     model = tf.keras.Model(inputs, outputs)
-    model.compile('sgd', loss=tf.keras.losses.GIoULoss())
+    model.compile('sgd', loss=tfa.losses.GIoULoss())
     ```
 
     Args:
-      mode: one of ['giou', 'iou'], decided to calculate giou loss or iou loss.
+      mode: one of ['giou', 'iou'], decided to calculate GIoU or IoU loss.
     """
 
     def __init__(self,
@@ -76,8 +78,7 @@ def giou_loss(y_true, y_pred, mode='giou'):
         box in boxes are encoded as [y_min, x_min, y_max, x_max].
         y_pred: predictions tensor. The coordinates of the each bounding
         box in boxes are encoded as [y_min, x_min, y_max, x_max].
-        mode: one of ['giou', 'iou'],
-        decided to calculate giou loss or iou loss.
+        mode: one of ['giou', 'iou'], decided to calculate GIoU or IoU loss.
 
     Returns:
         GIoU loss float `Tensor`.
@@ -100,8 +101,7 @@ def do_giou_calculate(b1, b2, mode='giou'):
         encoded as [y_min, x_min, y_max, x_max].
         b2: the other bounding box. The coordinates of the each bounding box
         in boxes are encoded as [y_min, x_min, y_max, x_max].
-        mode: one of ['giou', 'iou'],
-        decided to calculate giou loss or iou loss.
+        mode: one of ['giou', 'iou'], decided to calculate GIoU or IoU loss.
 
     Returns:
         GIoU loss float `Tensor`.

From e42ccd54d6331b593a224cbe05350f19ad289eab Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Tue, 19 Nov 2019 11:28:38 +0000
Subject: [PATCH 19/22] make private

---
 tensorflow_addons/losses/giou_loss.py | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index 0de33ea88c..aca70e2268 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -83,18 +83,17 @@ def giou_loss(y_true, y_pred, mode='giou'):
     Returns:
         GIoU loss float `Tensor`.
     """
-    # if y_true or y_pred:
     if mode not in ['giou', 'iou']:
         raise ValueError("Value of mode should be 'iou' or 'giou'")
     y_pred = tf.convert_to_tensor(y_pred)
     y_true = tf.cast(y_true, y_pred.dtype)
-    giou = do_giou_calculate(y_pred, y_true, mode)
+    giou = _calculate_giou(y_pred, y_true, mode)
 
     # compute the final loss and return
     return 1 - giou
 
 
-def do_giou_calculate(b1, b2, mode='giou'):
+def _calculate_giou(b1, b2, mode='giou'):
     """
     Args:
         b1: bounding box. The coordinates of the each bounding box in boxes are

From a03a320a48bb7890e4b77f9da8e4fc3a9caabcd6 Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Wed, 20 Nov 2019 07:16:45 +0000
Subject: [PATCH 20/22] add interger test

---
 tensorflow_addons/losses/README.md         |  2 +-
 tensorflow_addons/losses/giou_loss.py      | 13 +++++++------
 tensorflow_addons/losses/giou_loss_test.py | 12 +++++++++++-
 3 files changed, 19 insertions(+), 8 deletions(-)

diff --git a/tensorflow_addons/losses/README.md b/tensorflow_addons/losses/README.md
index ec570a51a8..2b40ad2f5e 100644
--- a/tensorflow_addons/losses/README.md
+++ b/tensorflow_addons/losses/README.md
@@ -5,7 +5,7 @@
 |:---------- |:----------- |:------------- |
 | contrastive |  @WindQAQ | windqaq@gmail.com |
 | focal_loss | @SSaishruthi  | saishruthi.tn@gmail.com |
-| giou_loss | @who who who  | fsx950223@gmail.com |
+| giou_loss | @fsx950223  | fsx950223@gmail.com |
 | lifted | @rahulunair | rahulunair@gmail.com  |
 | npairs | @WindQAQ | windqaq@gmail.com |
 | sparsemax_loss | @AndreasMadsen | amwwebdk+github@gmail.com |
diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index aca70e2268..b7743ac596 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -29,7 +29,7 @@ class GIoULoss(tf.keras.losses.Loss):
     [Generalized Intersection over Union:
     A Metric and A Loss for Bounding Box Regression]
     (https://giou.stanford.edu/GIoU.pdf).
-    GIoU is a enhance for model which use IoU in object detection.
+    GIoU is an enhancement for models which use IoU in object detection.
 
     Usage:
 
@@ -75,9 +75,9 @@ def giou_loss(y_true, y_pred, mode='giou'):
     """
     Args:
         y_true: true targets tensor. The coordinates of the each bounding
-        box in boxes are encoded as [y_min, x_min, y_max, x_max].
+            box in boxes are encoded as [y_min, x_min, y_max, x_max].
         y_pred: predictions tensor. The coordinates of the each bounding
-        box in boxes are encoded as [y_min, x_min, y_max, x_max].
+            box in boxes are encoded as [y_min, x_min, y_max, x_max].
         mode: one of ['giou', 'iou'], decided to calculate GIoU or IoU loss.
 
     Returns:
@@ -86,10 +86,11 @@ def giou_loss(y_true, y_pred, mode='giou'):
     if mode not in ['giou', 'iou']:
         raise ValueError("Value of mode should be 'iou' or 'giou'")
     y_pred = tf.convert_to_tensor(y_pred)
+    if y_pred.dtype.is_floating is not True:
+        y_pred=tf.cast(y_pred,tf.float32)
     y_true = tf.cast(y_true, y_pred.dtype)
     giou = _calculate_giou(y_pred, y_true, mode)
 
-    # compute the final loss and return
     return 1 - giou
 
 
@@ -97,9 +98,9 @@ def _calculate_giou(b1, b2, mode='giou'):
     """
     Args:
         b1: bounding box. The coordinates of the each bounding box in boxes are
-        encoded as [y_min, x_min, y_max, x_max].
+            encoded as [y_min, x_min, y_max, x_max].
         b2: the other bounding box. The coordinates of the each bounding box
-        in boxes are encoded as [y_min, x_min, y_max, x_max].
+            in boxes are encoded as [y_min, x_min, y_max, x_max].
         mode: one of ['giou', 'iou'], decided to calculate GIoU or IoU loss.
 
     Returns:
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
index e950079fe5..129d6daf97 100644
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -44,7 +44,7 @@ def test_iou(self, dtype):
                              dtype=dtype)
         boxes2 = tf.constant([[3.0, 4.0, 6.0, 8.0], [14.0, 14.0, 15.0, 15.0]],
                              dtype=dtype)
-        expected_result = tf.constant([14.0 / 16.0, 1.], dtype=dtype)
+        expected_result = tf.constant([0.875, 1.], dtype=dtype)
         loss = giou_loss(boxes1, boxes2, mode='iou')
         self.assertAllCloseAccordingToType(loss, expected_result)
 
@@ -61,6 +61,16 @@ def test_giou_loss(self, dtype):
         loss = giou_loss(boxes1, boxes2)
         self.assertAllCloseAccordingToType(loss, expected_result)
 
+    def test_with_integer(self):
+        boxes1 = tf.constant([[4, 3, 7, 5], [5, 6, 10, 7]],
+                             dtype=tf.int32)
+        boxes2 = tf.constant([[3, 4, 6, 8], [14, 14, 15, 15]],
+                             dtype=tf.int32)
+        expected_result = tf.constant(
+            [1.07500000298023224, 1.9333333373069763], dtype=tf.float32)
+        loss = giou_loss(boxes1, boxes2)
+        self.assertAllCloseAccordingToType(loss, expected_result)
+
     @parameterized.named_parameters(("float16", np.float16),
                                     ("float32", np.float32),
                                     ("float64", np.float64))

From 4b915d1b63ba6c7dcffd168bbba87ce5d7744abc Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Wed, 20 Nov 2019 07:17:55 +0000
Subject: [PATCH 21/22] format code

---
 tensorflow_addons/losses/giou_loss.py      | 2 +-
 tensorflow_addons/losses/giou_loss_test.py | 6 ++----
 2 files changed, 3 insertions(+), 5 deletions(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index b7743ac596..d04fc42037 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -87,7 +87,7 @@ def giou_loss(y_true, y_pred, mode='giou'):
         raise ValueError("Value of mode should be 'iou' or 'giou'")
     y_pred = tf.convert_to_tensor(y_pred)
     if y_pred.dtype.is_floating is not True:
-        y_pred=tf.cast(y_pred,tf.float32)
+        y_pred = tf.cast(y_pred, tf.float32)
     y_true = tf.cast(y_true, y_pred.dtype)
     giou = _calculate_giou(y_pred, y_true, mode)
 
diff --git a/tensorflow_addons/losses/giou_loss_test.py b/tensorflow_addons/losses/giou_loss_test.py
index 129d6daf97..4070d526b0 100644
--- a/tensorflow_addons/losses/giou_loss_test.py
+++ b/tensorflow_addons/losses/giou_loss_test.py
@@ -62,10 +62,8 @@ def test_giou_loss(self, dtype):
         self.assertAllCloseAccordingToType(loss, expected_result)
 
     def test_with_integer(self):
-        boxes1 = tf.constant([[4, 3, 7, 5], [5, 6, 10, 7]],
-                             dtype=tf.int32)
-        boxes2 = tf.constant([[3, 4, 6, 8], [14, 14, 15, 15]],
-                             dtype=tf.int32)
+        boxes1 = tf.constant([[4, 3, 7, 5], [5, 6, 10, 7]], dtype=tf.int32)
+        boxes2 = tf.constant([[3, 4, 6, 8], [14, 14, 15, 15]], dtype=tf.int32)
         expected_result = tf.constant(
             [1.07500000298023224, 1.9333333373069763], dtype=tf.float32)
         loss = giou_loss(boxes1, boxes2)

From f8ae33c5a1450ea4eca5ed3bbeddf35db76f34ee Mon Sep 17 00:00:00 2001
From: fsx950223 <fsx950223@gmail.com>
Date: Wed, 20 Nov 2019 09:35:32 +0000
Subject: [PATCH 22/22] change expression

---
 tensorflow_addons/losses/giou_loss.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tensorflow_addons/losses/giou_loss.py b/tensorflow_addons/losses/giou_loss.py
index d04fc42037..e864528ec2 100644
--- a/tensorflow_addons/losses/giou_loss.py
+++ b/tensorflow_addons/losses/giou_loss.py
@@ -86,7 +86,7 @@ def giou_loss(y_true, y_pred, mode='giou'):
     if mode not in ['giou', 'iou']:
         raise ValueError("Value of mode should be 'iou' or 'giou'")
     y_pred = tf.convert_to_tensor(y_pred)
-    if y_pred.dtype.is_floating is not True:
+    if not y_pred.dtype.is_floating:
         y_pred = tf.cast(y_pred, tf.float32)
     y_true = tf.cast(y_true, y_pred.dtype)
     giou = _calculate_giou(y_pred, y_true, mode)