Skip to content

Commit ad70fc4

Browse files
azkalot1yuki-inaho
azkalot1
authored and
yuki-inaho
committed
Add regnet, res2net, resnest, sknet encoders (qubvel-org#286)
* add regnet, res2net, resnest, sknet encoders * fixed tests * add raise erorr for dilated encoders in deeplabv3 and PAN * raise correct error for encoders without dilation * update README * fix README
1 parent b969ee7 commit ad70fc4

File tree

8 files changed

+861
-5
lines changed

8 files changed

+861
-5
lines changed

‎README.md‎

Lines changed: 42 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -140,6 +140,48 @@ Congratulations! You are done! Now you can train your model with your favorite f
140140
|timm-efficientnet-b7 |imagenet / advprop / noisy-student|63M |
141141
|timm-efficientnet-b8 |imagenet / advprop |84M |
142142
|timm-efficientnet-l2 |noisy-student |474M |
143+
|timm-resnest14d |imagenet |8M |
144+
|timm-resnest26d |imagenet |15M |
145+
|timm-resnest50d |imagenet |25M |
146+
|timm-resnest101e |imagenet |46M |
147+
|timm-resnest200e |imagenet |68M |
148+
|timm-resnest269e |imagenet |108M |
149+
|timm-resnest50d_4s2x40d |imagenet |28M |
150+
|timm-resnest50d_1s4x24d |imagenet |23M |
151+
|timm-res2net50_26w_4s |imagenet |23M |
152+
|timm-res2net101_26w_4s |imagenet |43M |
153+
|timm-res2net50_26w_6s |imagenet |35M |
154+
|timm-res2net50_26w_8s |imagenet |46M |
155+
|timm-res2net50_48w_2s |imagenet |23M |
156+
|timm-res2net50_14w_8s |imagenet |23M |
157+
|timm-res2next50 |imagenet |22M |
158+
|timm-regnetx_002 |imagenet |2M |
159+
|timm-regnetx_004 |imagenet |4M |
160+
|timm-regnetx_006 |imagenet |5M |
161+
|timm-regnetx_008 |imagenet |6M |
162+
|timm-regnetx_016 |imagenet |8M |
163+
|timm-regnetx_032 |imagenet |14M |
164+
|timm-regnetx_040 |imagenet |20M |
165+
|timm-regnetx_064 |imagenet |24M |
166+
|timm-regnetx_080 |imagenet |37M |
167+
|timm-regnetx_120 |imagenet |43M |
168+
|timm-regnetx_160 |imagenet |52M |
169+
|timm-regnetx_320 |imagenet |105M |
170+
|timm-regnety_002 |imagenet |2M |
171+
|timm-regnety_004 |imagenet |3M |
172+
|timm-regnety_006 |imagenet |5M |
173+
|timm-regnety_008 |imagenet |5M |
174+
|timm-regnety_016 |imagenet |10M |
175+
|timm-regnety_032 |imagenet |17M |
176+
|timm-regnety_040 |imagenet |19M |
177+
|timm-regnety_064 |imagenet |29M |
178+
|timm-regnety_080 |imagenet |37M |
179+
|timm-regnety_120 |imagenet |49M |
180+
|timm-regnety_160 |imagenet |80M |
181+
|timm-regnety_320 |imagenet |141M |
182+
|timm-skresnet18 |imagenet |11M |
183+
|timm-skresnet34 |imagenet |21M |
184+
|timm-skresnext50_32x4d |imagenet |25M |
143185

144186
\* `ssl`, `wsl` - semi-supervised and weakly-supervised learning on ImageNet ([repo](https://github.com/facebookresearch/semi-supervised-ImageNet1K-models)).
145187

‎segmentation_models_pytorch/deeplabv3/model.py‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -34,7 +34,7 @@ class DeepLabV3(SegmentationModel):
3434
.. _DeeplabV3:
3535
https://arxiv.org/abs/1706.05587
3636
"""
37-
37+
3838
def __init__(
3939
self,
4040
encoder_name: str = "resnet34",

‎segmentation_models_pytorch/encoders/__init__.py‎

Lines changed: 8 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -12,7 +12,10 @@
1212
from .mobilenet import mobilenet_encoders
1313
from .xception import xception_encoders
1414
from .timm_efficientnet import timm_efficientnet_encoders
15-
15+
from .timm_resnest import timm_resnest_encoders
16+
from .timm_res2net import timm_res2net_encoders
17+
from .timm_regnet import timm_regnet_encoders
18+
from .timm_sknet import timm_sknet_encoders
1619
from ._preprocessing import preprocess_input
1720

1821
encoders = {}
@@ -27,6 +30,10 @@
2730
encoders.update(mobilenet_encoders)
2831
encoders.update(xception_encoders)
2932
encoders.update(timm_efficientnet_encoders)
33+
encoders.update(timm_resnest_encoders)
34+
encoders.update(timm_res2net_encoders)
35+
encoders.update(timm_regnet_encoders)
36+
encoders.update(timm_sknet_encoders)
3037

3138

3239
def get_encoder(name, in_channels=3, depth=5, weights=None):

0 commit comments

Comments
 (0)