mirrors
/
mmsegmentation
mirror of https://github.com/open-mmlab/mmsegmentation.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

[Feature] Add OS16 DeepLab (#154) 

* Add D16-MG124 models

* Use MMCV DepthSepConv

* add OHEM

* add warmup

* fixed test

* fixed test

* change to bs 16

* revert config

* add models
pull/169/head
Jerry Jiarui XU 2020-09-29 20:00:48 +08:00 committed by GitHub
parent 65da3e3752
commit e1f4f51dc6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
6 changed files with 51 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
configs/deeplabv3/README.md
View File

@ -25,6 +25,8 @@ Note: `D-8` here corresponding to the output stride 8 setting for DeepLab series
| DeepLabV3 | R-101-D8 | 512x1024 | 80000 | - | - | 80.20 | 81.21 | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_512x1024_80k_cityscapes/deeplabv3_r101-d8_512x1024_80k_cityscapes_20200606_113503-9e428899.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_512x1024_80k_cityscapes/deeplabv3_r101-d8_512x1024_80k_cityscapes_20200606_113503.log.json) |
| DeepLabV3 | R-50-D8 | 769x769 | 80000 | - | - | 79.89 | 81.06 | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r50-d8_769x769_80k_cityscapes/deeplabv3_r50-d8_769x769_80k_cityscapes_20200606_221338-788d6228.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r50-d8_769x769_80k_cityscapes/deeplabv3_r50-d8_769x769_80k_cityscapes_20200606_221338.log.json) |
| DeepLabV3 | R-101-D8 | 769x769 | 80000 | - | - | 79.67 | 80.81 | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_769x769_80k_cityscapes/deeplabv3_r101-d8_769x769_80k_cityscapes_20200607_013353-60e95418.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_769x769_80k_cityscapes/deeplabv3_r101-d8_769x769_80k_cityscapes_20200607_013353.log.json) |
| DeepLabV3 | R-101-D16-MG124 | 512x1024 | 40000 | 4.7 | - 6.96 | 76.71 | 78.63 | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d16-mg124_512x1024_40k_cityscapes/deeplabv3_r101-d16-mg124_512x1024_40k_cityscapes_20200908_005644-67b0c992.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d16-mg124_512x1024_40k_cityscapes/deeplabv3_r101-d16-mg124_512x1024_40k_cityscapes-20200908_005644.log.json) |
| DeepLabV3 | R-101-D16-MG124 | 512x1024 | 80000 | - | - | 78.36 | 79.84 | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d16-mg124_512x1024_80k_cityscapes/deeplabv3_r101-d16-mg124_512x1024_80k_cityscapes_20200908_005644-57bb8425.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d16-mg124_512x1024_80k_cityscapes/deeplabv3_r101-d16-mg124_512x1024_80k_cityscapes-20200908_005644.log.json) |
### ADE20K
| Method | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | download |

11
configs/deeplabv3/deeplabv3_r101-d16-mg124_512x1024_40k_cityscapes.py 100644
View File

@ -0,0 +1,11 @@
_base_ = './deeplabv3_r50-d8_512x1024_40k_cityscapes.py'
model = dict(
pretrained='open-mmlab://resnet101_v1c',
backbone=dict(
depth=101,
dilations=(1, 1, 1, 2),
strides=(1, 2, 2, 1),
multi_grid=(1, 2, 4)),
decode_head=dict(
dilations=(1, 6, 12, 18),
sampler=dict(type='OHEMPixelSampler', min_kept=100000)))

11
configs/deeplabv3/deeplabv3_r101-d16-mg124_512x1024_80k_cityscapes.py 100644
View File

@ -0,0 +1,11 @@
_base_ = './deeplabv3_r50-d8_512x1024_80k_cityscapes.py'
model = dict(
pretrained='open-mmlab://resnet101_v1c',
backbone=dict(
depth=101,
dilations=(1, 1, 1, 2),
strides=(1, 2, 2, 1),
multi_grid=(1, 2, 4)),
decode_head=dict(
dilations=(1, 6, 12, 18),
sampler=dict(type='OHEMPixelSampler', min_kept=100000)))

6
configs/deeplabv3plus/README.md
View File

@ -12,7 +12,9 @@
## Results and models
Note: `D-8` here corresponding to the output stride 8 setting for DeepLab series.
Note:
`D-8`/`D-16` here corresponding to the output stride 8/16 setting for DeepLab series.
`MG-124` stands for multi-grid dilation in the last stage of ResNet.
### Cityscapes
| Method | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | download |
@ -25,6 +27,8 @@ Note: `D-8` here corresponding to the output stride 8 setting for DeepLab series
| DeepLabV3+ | R-101-D8 | 512x1024 | 80000 | - | - | 80.97 | 82.03 | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d8_512x1024_80k_cityscapes/deeplabv3plus_r101-d8_512x1024_80k_cityscapes_20200606_114143-068fcfe9.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d8_512x1024_80k_cityscapes/deeplabv3plus_r101-d8_512x1024_80k_cityscapes_20200606_114143.log.json) |
| DeepLabV3+ | R-50-D8 | 769x769 | 80000 | - | - | 79.83 | 81.48 | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r50-d8_769x769_80k_cityscapes/deeplabv3plus_r50-d8_769x769_80k_cityscapes_20200606_210233-0e9dfdc4.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r50-d8_769x769_80k_cityscapes/deeplabv3plus_r50-d8_769x769_80k_cityscapes_20200606_210233.log.json) |
| DeepLabV3+ | R-101-D8 | 769x769 | 80000 | - | - | 80.98 | 82.18 | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d8_769x769_80k_cityscapes/deeplabv3plus_r101-d8_769x769_80k_cityscapes_20200607_000405-a7573d20.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d8_769x769_80k_cityscapes/deeplabv3plus_r101-d8_769x769_80k_cityscapes_20200607_000405.log.json) |
| DeepLabV3+ | R-101-D16-MG124 | 512x1024 | 40000 | 5.8 | 7.48 | 79.09 | 80.36 | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d16-mg124_512x1024_40k_cityscapes/deeplabv3plus_r101-d16-mg124_512x1024_40k_cityscapes_20200908_005644-cf9ce186.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d16-mg124_512x1024_40k_cityscapes/deeplabv3plus_r101-d16-mg124_512x1024_40k_cityscapes-20200908_005644.log.json) |
| DeepLabV3+ | R-101-D16-MG124 | 512x1024 | 80000 | 9.9 | - | 79.90 | 81.33 | [model](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d16-mg124_512x1024_80k_cityscapes/deeplabv3plus_r101-d16-mg124_512x1024_80k_cityscapes_20200908_005644-ee6158e0.pth) | [log](https://download.openmmlab.com/mmsegmentation/v0.5/deeplabv3plus/deeplabv3plus_r101-d16-mg124_512x1024_80k_cityscapes/deeplabv3plus_r101-d16-mg124_512x1024_80k_cityscapes-20200908_005644.log.json) |
### ADE20K
| Method | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | download |

11
configs/deeplabv3plus/deeplabv3plus_r101-d16-mg124_512x1024_40k_cityscapes.py 100644
View File

@ -0,0 +1,11 @@
_base_ = './deeplabv3plus_r50-d8_512x1024_40k_cityscapes.py'
model = dict(
pretrained='open-mmlab://resnet101_v1c',
backbone=dict(
depth=101,
dilations=(1, 1, 1, 2),
strides=(1, 2, 2, 1),
multi_grid=(1, 2, 4)),
decode_head=dict(
dilations=(1, 6, 12, 18),
sampler=dict(type='OHEMPixelSampler', min_kept=100000)))

11
configs/deeplabv3plus/deeplabv3plus_r101-d16-mg124_512x1024_80k_cityscapes.py 100644
View File

@ -0,0 +1,11 @@
_base_ = './deeplabv3plus_r50-d8_512x1024_80k_cityscapes.py'
model = dict(
pretrained='open-mmlab://resnet101_v1c',
backbone=dict(
depth=101,
dilations=(1, 1, 1, 2),
strides=(1, 2, 2, 1),
multi_grid=(1, 2, 4)),
decode_head=dict(
dilations=(1, 6, 12, 18),
sampler=dict(type='OHEMPixelSampler', min_kept=100000)))