[Benchmark] Uploading FastFCN on ADE20K (#972)

* Uploading FastFCN on ADE20K

* fixing lint error

configs/fastfcn/README.md

@@ -35,6 +35,17 @@ year={2019}
 | EncNet + JPU | R-50-D32  | 512x1024  |   80000 | 8.15        |  4.77             | 77.97 |79.92          | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/fastfcn/fastfcn_r50-d32_jpu_enc_512x1024_80k_cityscapes.py)   | [model](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_512x1024_80k_cityscapes/fastfcn_r50-d32_jpu_enc_512x1024_80k_cityscapes_20210928_030036-78da5046.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_512x1024_80k_cityscapes/fastfcn_r50-d32_jpu_enc_512x1024_80k_cityscapes_20210928_030036.log.json)    |
 | EncNet + JPU (4x4)| R-50-D32  | 512x1024  |   80000 | 15.45        | -              | 78.6 |         80.25 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/fastfcn/fastfcn_r50-d32_jpu_enc_4x4_512x1024_80k_cityscapes.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_4x4_512x1024_80k_cityscapes/fastfcn_r50-d32_jpu_enc_4x4_512x1024_80k_cityscapes_20210926_093217-e1eb6dbb.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_4x4_512x1024_80k_cityscapes/fastfcn_r50-d32_jpu_enc_4x4_512x1024_80k_cityscapes_20210926_093217.log.json)     |
 
+### ADE20K
+
+| Method    | Backbone  | Crop Size | Lr schd | Mem (GB) | Inf time (fps) |  mIoU | mIoU(ms+flip) | config                                                                                  | download                                                                                                                                                                                                                                                       |
+| --------- | --------- | --------- | ------: | -------- | -------------- | ----: | ------------- | --------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| DeepLabV3 + JPU | R-50-D32 | 512x1024  | 80000 | 8.46 | 12.06 | 41.88 | 42.91 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k/fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k_20211013_190619-3aa40f2d.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k/fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k_20211013_190619.log.json)  |
+| DeepLabV3 + JPU | R-50-D32 | 512x1024  | 160000 | - | - | 43.58 | 44.92 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_160k_ade20k.py) | [model](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_160k_ade20k/fastfcn_r50-d32_jpu_aspp_512x512_160k_ade20k_20211008_152246-27036aee.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_160k_ade20k/fastfcn_r50-d32_jpu_aspp_512x512_160k_ade20k_20211008_152246.log.json) |
+| PSPNet + JPU | R-50-D32   | 512x1024  |   80000 | 8.02        |  19.21             | 41.40 |   42.12       | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k.py)  |  [model](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k/fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k_20210930_225137-993d07c8.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k/fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k_20210930_225137.log.json)       |
+| PSPNet + JPU | R-50-D32   | 512x1024  |   160000 |  -       |  -             | 42.63 |   43.71       | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k.py)   | [model](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k/fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k_20211008_105455-e8f5a2fd.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k/fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k_20211008_105455.log.json)         |
+| EncNet + JPU | R-50-D32  | 512x1024  |   80000 | 9.67        |  17.23             | 40.88 | 42.36          | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_80k_ade20k.py)   | [model](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_80k_ade20k/fastfcn_r50-d32_jpu_enc_512x512_80k_ade20k_20210930_225214-65aef6dd.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_80k_ade20k/fastfcn_r50-d32_jpu_enc_512x512_80k_ade20k_20210930_225214.log.json)    |
+| EncNet + JPU | R-50-D32  | 512x1024  |   160000 | -        | -              | 42.50 |         44.21 | [config](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_160k_ade20k.py)  | [model](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_160k_ade20k/fastfcn_r50-d32_jpu_enc_512x512_160k_ade20k_20211008_105456-d875ce3c.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_160k_ade20k/fastfcn_r50-d32_jpu_enc_512x512_160k_ade20k_20211008_105456.log.json)     |
+
 Note:
 
 - `4x4` means 4 GPUs with 4 samples per GPU in training, default setting is 4 GPUs with 2 samples per GPU in training.

configs/fastfcn/fastfcn.yml

@@ -3,6 +3,7 @@ Collections:
   Metadata:
     Training Data:
     - Cityscapes
+    - ADE20K
   Paper:
     URL: https://arxiv.org/abs/1903.11816
     Title: 'FastFCN: Rethinking Dilated Convolution in the Backbone for Semantic Segmentation'
@@ -124,3 +125,111 @@ Models:
       mIoU(ms+flip): 80.25
   Config: configs/fastfcn/fastfcn_r50-d32_jpu_enc_4x4_512x1024_80k_cityscapes.py
   Weights: https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_4x4_512x1024_80k_cityscapes/fastfcn_r50-d32_jpu_enc_4x4_512x1024_80k_cityscapes_20210926_093217-e1eb6dbb.pth
+- Name: fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k
+  In Collection: fastfcn
+  Metadata:
+    backbone: R-50-D32
+    crop size: (512,1024)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 82.92
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,1024)
+    memory (GB): 8.46
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: ADE20K
+    Metrics:
+      mIoU: 41.88
+      mIoU(ms+flip): 42.91
+  Config: configs/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k/fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k_20211013_190619-3aa40f2d.pth
+- Name: fastfcn_r50-d32_jpu_aspp_512x512_160k_ade20k
+  In Collection: fastfcn
+  Metadata:
+    backbone: R-50-D32
+    crop size: (512,1024)
+    lr schd: 160000
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: ADE20K
+    Metrics:
+      mIoU: 43.58
+      mIoU(ms+flip): 44.92
+  Config: configs/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_160k_ade20k.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_160k_ade20k/fastfcn_r50-d32_jpu_aspp_512x512_160k_ade20k_20211008_152246-27036aee.pth
+- Name: fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k
+  In Collection: fastfcn
+  Metadata:
+    backbone: R-50-D32
+    crop size: (512,1024)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 52.06
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,1024)
+    memory (GB): 8.02
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: ADE20K
+    Metrics:
+      mIoU: 41.4
+      mIoU(ms+flip): 42.12
+  Config: configs/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k/fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k_20210930_225137-993d07c8.pth
+- Name: fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k
+  In Collection: fastfcn
+  Metadata:
+    backbone: R-50-D32
+    crop size: (512,1024)
+    lr schd: 160000
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: ADE20K
+    Metrics:
+      mIoU: 42.63
+      mIoU(ms+flip): 43.71
+  Config: configs/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k/fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k_20211008_105455-e8f5a2fd.pth
+- Name: fastfcn_r50-d32_jpu_enc_512x512_80k_ade20k
+  In Collection: fastfcn
+  Metadata:
+    backbone: R-50-D32
+    crop size: (512,1024)
+    lr schd: 80000
+    inference time (ms/im):
+    - value: 58.04
+      hardware: V100
+      backend: PyTorch
+      batch size: 1
+      mode: FP32
+      resolution: (512,1024)
+    memory (GB): 9.67
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: ADE20K
+    Metrics:
+      mIoU: 40.88
+      mIoU(ms+flip): 42.36
+  Config: configs/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_80k_ade20k.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_80k_ade20k/fastfcn_r50-d32_jpu_enc_512x512_80k_ade20k_20210930_225214-65aef6dd.pth
+- Name: fastfcn_r50-d32_jpu_enc_512x512_160k_ade20k
+  In Collection: fastfcn
+  Metadata:
+    backbone: R-50-D32
+    crop size: (512,1024)
+    lr schd: 160000
+  Results:
+  - Task: Semantic Segmentation
+    Dataset: ADE20K
+    Metrics:
+      mIoU: 42.5
+      mIoU(ms+flip): 44.21
+  Config: configs/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_160k_ade20k.py
+  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_160k_ade20k/fastfcn_r50-d32_jpu_enc_512x512_160k_ade20k_20211008_105456-d875ce3c.pth

configs/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_160k_ade20k.py

@@ -0,0 +1,20 @@
+# model settings
+_base_ = './fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k.py'
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    decode_head=dict(
+        _delete_=True,
+        type='ASPPHead',
+        in_channels=2048,
+        in_index=2,
+        channels=512,
+        dilations=(1, 12, 24, 36),
+        dropout_ratio=0.1,
+        num_classes=150,
+        norm_cfg=norm_cfg,
+        align_corners=False,
+        loss_decode=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
+    # model training and testing settings
+    train_cfg=dict(),
+    test_cfg=dict(mode='whole'))

configs/fastfcn/fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k.py

@@ -0,0 +1,20 @@
+# model settings
+_base_ = './fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k.py'
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    decode_head=dict(
+        _delete_=True,
+        type='ASPPHead',
+        in_channels=2048,
+        in_index=2,
+        channels=512,
+        dilations=(1, 12, 24, 36),
+        dropout_ratio=0.1,
+        num_classes=150,
+        norm_cfg=norm_cfg,
+        align_corners=False,
+        loss_decode=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
+    # model training and testing settings
+    train_cfg=dict(),
+    test_cfg=dict(mode='whole'))

configs/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_160k_ade20k.py

@@ -0,0 +1,24 @@
+# model settings
+_base_ = './fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k.py'
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    decode_head=dict(
+        _delete_=True,
+        type='EncHead',
+        in_channels=[512, 1024, 2048],
+        in_index=(0, 1, 2),
+        channels=512,
+        num_codes=32,
+        use_se_loss=True,
+        add_lateral=False,
+        dropout_ratio=0.1,
+        num_classes=150,
+        norm_cfg=norm_cfg,
+        align_corners=False,
+        loss_decode=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_se_decode=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=0.2)),
+    # model training and testing settings
+    train_cfg=dict(),
+    test_cfg=dict(mode='whole'))

configs/fastfcn/fastfcn_r50-d32_jpu_enc_512x512_80k_ade20k.py

@@ -0,0 +1,24 @@
+# model settings
+_base_ = './fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k.py'
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    decode_head=dict(
+        _delete_=True,
+        type='EncHead',
+        in_channels=[512, 1024, 2048],
+        in_index=(0, 1, 2),
+        channels=512,
+        num_codes=32,
+        use_se_loss=True,
+        add_lateral=False,
+        dropout_ratio=0.1,
+        num_classes=150,
+        norm_cfg=norm_cfg,
+        align_corners=False,
+        loss_decode=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_se_decode=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=0.2)),
+    # model training and testing settings
+    train_cfg=dict(),
+    test_cfg=dict(mode='whole'))

configs/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_160k_ade20k.py

@@ -0,0 +1,7 @@
+_base_ = [
+    '../_base_/models/fastfcn_r50-d32_jpu_psp.py',
+    '../_base_/datasets/ade20k.py', '../_base_/default_runtime.py',
+    '../_base_/schedules/schedule_160k.py'
+]
+model = dict(
+    decode_head=dict(num_classes=150), auxiliary_head=dict(num_classes=150))

configs/fastfcn/fastfcn_r50-d32_jpu_psp_512x512_80k_ade20k.py

@@ -0,0 +1,7 @@
+_base_ = [
+    '../_base_/models/fastfcn_r50-d32_jpu_psp.py',
+    '../_base_/datasets/ade20k.py', '../_base_/default_runtime.py',
+    '../_base_/schedules/schedule_80k.py'
+]
+model = dict(
+    decode_head=dict(num_classes=150), auxiliary_head=dict(num_classes=150))