add more results of OCRNet (#20)

* update the HRNet-OCR & add ResNet-101-OCR

* revise the script

* add the results of resnet-101+ocr

* add cascade ocr, aspp ocr

* add comparison table

* move comparison table

* support ocr+decoder

* revise the ocrnet_sep_aspp

* update the results of ocrnet

* update the results of ocrnet

* add sep-ocr-variants

* add bs2x exp of deeplabv3/v3+

* apply sep-conv in ocr module

* update the results

* update the results

* update the results of OCRNet

* update the results of OCRNet

* correct the results

* verify the release branch

* init the release branch

* add more results of ocrnet and ocrnetplus

* resolve the conflicts

* rename OCRNetPlus as OCRNet+

* fix the format

* fix the lint issues

* fix the lint issues

* fix the lint issues

* fix the lint isort issues

* fix the lint yapf issues

* fix the format issues

* remove the changes by the master branch

* remove the changes by the master branch

* remove the changes by the master branch

* remove the changes by the master branch

* remove the changes by the master branch

* add the logs folder to .gitignore

* recover .gitignore

* update readme

* update readme

* reset the cudnn_benchmark

* revise the README of OCRNet

* revise the name

* revise the reference of OCRNet

* revise the Figure of OCRNet+

* update the results of OCR/OCR+

* update the results of OCR/OCR+

* update the results of OCR/OCR+

* fix the format issue

* fix the format issue

* remove the ocr+

* update the results

* update the results

* fix the conflicts

* fix the lint issue

* fix the lint issue

* fix the lint issue

* fix the lint issue

* fix the inconsistency

* add urls to README

* clean the code

* remove the schedule configs

* clean the custom code

* clean up

* remove ocr.png

Co-authored-by: Jiarui XU <xvjiarui0826@gmail.com>
pull/74/head
RainbowSecret 2020-08-13 22:16:27 -07:00 committed by GitHub
parent 9ba22456d9
commit 381eacb9a5
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 96 additions and 4 deletions

View File

@ -0,0 +1,47 @@
# model settings
norm_cfg = dict(type='SyncBN', requires_grad=True)
model = dict(
type='CascadeEncoderDecoder',
num_stages=2,
pretrained='open-mmlab://resnet50_v1c',
backbone=dict(
type='ResNetV1c',
depth=50,
num_stages=4,
out_indices=(0, 1, 2, 3),
dilations=(1, 1, 2, 4),
strides=(1, 2, 1, 1),
norm_cfg=norm_cfg,
norm_eval=False,
style='pytorch',
contract_dilation=True),
decode_head=[
dict(
type='FCNHead',
in_channels=1024,
in_index=2,
channels=256,
num_convs=1,
concat_input=False,
drop_out_ratio=0.1,
num_classes=19,
norm_cfg=norm_cfg,
align_corners=False,
loss_decode=dict(
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=0.4)),
dict(
type='OCRHead',
in_channels=2048,
in_index=3,
channels=512,
ocr_channels=256,
drop_out_ratio=0.1,
num_classes=19,
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')

View File

@ -1,18 +1,28 @@
# Object-Contextual Representations for Semantic Segmentation
## Introduction
```
@article{yuan2019ocr,
@article{YuanW18,
title={Ocnet: Object context network for scene parsing},
author={Yuhui Yuan and Jingdong Wang},
booktitle={arXiv preprint arXiv:1809.00916},
year={2018}
}
@article{YuanCW20,
title={Object-Contextual Representations for Semantic Segmentation},
author={Yuan Yuhui and Chen Xilin and Wang Jingdong},
journal={arXiv preprint arXiv:1909.11065},
year={2019}
author={Yuhui Yuan and Xilin Chen and Jingdong Wang},
booktitle={ECCV},
year={2020}
}
```
## Results and models
### Cityscapes
#### HRNet backbone
| Method | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | download |
|--------|--------------------|-----------|--------:|----------|----------------|------:|--------------:|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| OCRNet | HRNetV2p-W18-Small | 512x1024 | 40000 | 3.5 | 10.45 | 74.30 | 75.95 | [model](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_hr18s_512x1024_40k_cityscapes/ocrnet_hr18s_512x1024_40k_cityscapes_20200601_033304-fa2436c2.pth) &#124; [log](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_hr18s_512x1024_40k_cityscapes/ocrnet_hr18s_512x1024_40k_cityscapes_20200601_033304.log.json) |
@ -25,6 +35,16 @@
| OCRNet | HRNetV2p-W18 | 512x1024 | 160000 | - | - | 79.47 | 80.91 | [model](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_hr18_512x1024_160k_cityscapes/ocrnet_hr18_512x1024_160k_cityscapes_20200602_191001-b9172d0c.pth) &#124; [log](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_hr18_512x1024_160k_cityscapes/ocrnet_hr18_512x1024_160k_cityscapes_20200602_191001.log.json) |
| OCRNet | HRNetV2p-W48 | 512x1024 | 160000 | - | - | 81.35 | 82.70 | [model](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_hr48_512x1024_160k_cityscapes/ocrnet_hr48_512x1024_160k_cityscapes_20200602_191037-dfbf1b0c.pth) &#124; [log](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_hr48_512x1024_160k_cityscapes/ocrnet_hr48_512x1024_160k_cityscapes_20200602_191037.log.json) |
#### ResNet backbone
| Method | Backbone | Crop Size | Batch Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | download |
|--------|--------------------|-----------|--------|----------|-----------|----------------|------|--------------:|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| OCRNet | R-101-D8 | 512x1024 | 8 | 40000 | - | - | 80.09 | - | [model](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_r101-d8_512x1024_40k_b8_cityscapes/ocrnet_r101-d8_512x1024_40k_b8_cityscapes-02ac0f13.pth) &#124; [log](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_r101-d8_512x1024_40k_b8_cityscapes/ocrnet_r101-d8_512x1024_40k_b8_cityscapes_20200717_110721.log.json) |
| OCRNet | R-101-D8 | 512x1024 | 16 | 40000 | 8.8 | 3.02 | 80.30 | - | [model](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_r101-d8_512x1024_40k_b16_cityscapes/ocrnet_r101-d8_512x1024_40k_b16_cityscapes-db500f80.pth) &#124; [log](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_r101-d8_512x1024_40k_b16_cityscapes/ocrnet_r101-d8_512x1024_40k_b16_cityscapes_20200723_193726.log.json) |
| OCRNet | R-101-D8 | 512x1024 | 16 | 80000 | 8.8 | 3.02 | 80.81 | - | [model](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_r101-d8_512x1024_80k_b16_cityscapes/ocrnet_r101-d8_512x1024_80k_b16_cityscapes-78688424.pth) &#124; [log](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/ocrnet/ocrnet_r101-d8_512x1024_80k_b16_cityscapes/ocrnet_r101-d8_512x1024_80k_b16_cityscapes_20200723_192421.log.json) |
### ADE20K
| Method | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | download |
|--------|--------------------|-----------|--------:|----------|----------------|------:|--------------:|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|

View File

@ -0,0 +1,9 @@
_base_ = [
'../_base_/models/ocrnet_r50-d8.py',
'../_base_/datasets/cityscapes.py',
'../_base_/default_runtime.py',
'../_base_/schedules/schedule_40k.py'
]
model = dict(pretrained='open-mmlab://resnet101_v1c', backbone=dict(depth=101))
optimizer = dict(lr=0.02)
lr_config = dict(min_lr=2e-4)

View File

@ -0,0 +1,7 @@
_base_ = [
'../_base_/models/ocrnet_r50-d8.py',
'../_base_/datasets/cityscapes.py',
'../_base_/default_runtime.py',
'../_base_/schedules/schedule_40k.py'
]
model = dict(pretrained='open-mmlab://resnet101_v1c', backbone=dict(depth=101))

View File

@ -0,0 +1,9 @@
_base_ = [
'../_base_/models/ocrnet_r50-d8.py',
'../_base_/datasets/cityscapes.py',
'../_base_/default_runtime.py',
'../_base_/schedules/schedule_80k.py'
]
model = dict(pretrained='open-mmlab://resnet101_v1c', backbone=dict(depth=101))
optimizer = dict(lr=0.02)
lr_config = dict(min_lr=2e-4)