Add Pascal Context to mmsegmentation #133

yamengxi · 2020-09-15T06:12:53Z

No description provided.

CLAassistant · 2020-09-15T06:13:01Z

All committers have signed the CLA.

codecov · 2020-09-15T06:22:46Z

Codecov Report

Merging #133 into master will increase coverage by 0.00%.
The diff coverage is 85.71%.

@@           Coverage Diff           @@
##           master     #133   +/-   ##
=======================================
  Coverage   83.03%   83.04%           
=======================================
  Files          84       85    +1     
  Lines        3968     3980   +12     
  Branches      625      627    +2     
=======================================
+ Hits         3295     3305   +10     
- Misses        539      541    +2     
  Partials      134      134

Flag	Coverage Δ
#unittests	`83.04% <85.71%> (+<0.01%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmseg/models/segmentors/encoder_decoder.py	`89.93% <ø> (-0.07%)`	⬇️
mmseg/datasets/pascal_context.py	`80.00% <80.00%> (ø)`
mmseg/core/evaluation/mean_iou.py	`100.00% <100.00%> (ø)`
mmseg/datasets/__init__.py	`100.00% <100.00%> (ø)`

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 768c3ee...1e73bf1. Read the comment docs.

xvjiarui · 2020-09-15T07:07:22Z

mmseg/__init__.py

@@ -2,7 +2,7 @@

 from .version import __version__, version_info

-MMCV_MIN = '1.1.2'
+MMCV_MIN = '1.1.0'


We may keep this file unchanged.

xvjiarui · 2020-09-15T07:07:50Z

mmseg/datasets/pascal_context.py

+ '''
+ PALETTE is testing
+ '''


We may remove this line if the palette works fine.

xvjiarui · 2020-09-15T07:08:13Z

mmseg/models/segmentors/encoder_decoder.py

@@ -91,7 +91,7 @@ def encode_decode(self, img, img_metas):
 size=img.shape[2:],
 mode='bilinear',
 align_corners=self.align_corners)
- return out
+ return out.exp()


Suggested change

return out.exp()

return out

xvjiarui · 2020-09-15T07:09:24Z

mmseg/models/segmentors/encoder_decoder.py

+ # assert h_crop <= h_img and w_crop <= w_img, (
+ #  'crop size should not greater than image size')


These lines could be removed. We may also add comments about what would happen if h_crop > h_img.

xvjiarui · 2020-09-15T07:09:41Z

run.sh

@@ -0,0 +1,15 @@
+GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=1 tools/slurm_train.sh mediaf pr_test configs/hrnet/fcn_hr48_480x480_40k_pascal_context.py --seed=0


This file is not necessary.

tools/convert_datasets/pascal_context.py

xvjiarui · 2020-09-15T07:10:54Z

tools/slurm_test.sh

@@ -21,4 +21,4 @@ srun -p ${PARTITION} \
 --cpus-per-task=${CPUS_PER_TASK} \
 --kill-on-bad-exit=1 \
 ${SRUN_ARGS} \
- python -u tools/test.py ${CONFIG} ${CHECKPOINT} --launcher="slurm" ${PY_ARGS}
+ python -u tools/test.py ${CONFIG} ${CHECKPOINT} --launcher="slurm" ${PY_ARGS} &


Unnecessary modification.

xvjiarui · 2020-09-15T07:11:04Z

tools/slurm_train.sh

@@ -20,4 +20,4 @@ srun -p ${PARTITION} \
 --cpus-per-task=${CPUS_PER_TASK} \
 --kill-on-bad-exit=1 \
 ${SRUN_ARGS} \
- python -u tools/train.py ${CONFIG} --launcher="slurm" ${PY_ARGS}
+ python -u tools/train.py ${CONFIG} --launcher="slurm" ${PY_ARGS} &


Unnecessary modification.

xvjiarui · 2020-09-15T07:11:22Z

tools/test.py

+ if 'CLASSES' not in checkpoint['meta']:
+ model.CLASSES = dataset.CLASSES
+ else:
+ model.CLASSES = checkpoint['meta']['CLASSES']
+
+ if 'PALETTE' not in checkpoint['meta']:
+ model.PALETTE = dataset.PALETTE
+ else:
+ model.PALETTE = checkpoint['meta']['PALETTE']


Unnecessary modification.

xvjiarui · 2020-09-15T07:12:22Z

configs/_base_/schedules/schedule_40k_lr4e-3.py

+# optimizer
+optimizer = dict(type='SGD', lr=0.004, momentum=0.9, weight_decay=0.0001)
+optimizer_config = dict()
+# learning policy
+lr_config = dict(policy='poly', power=0.9, min_lr=1e-4, by_epoch=False)
+# runtime settings
+total_iters = 40000
+checkpoint_config = dict(by_epoch=False, interval=4000)
+evaluation = dict(interval=4000, metric='mIoU')


We may delete this file and modify the corresponding lr directly.

xvjiarui · 2020-09-15T07:12:35Z

configs/_base_/schedules/schedule_80k_lr4e-3.py

+optimizer = dict(type='SGD', lr=0.004, momentum=0.9, weight_decay=0.0001)
+optimizer_config = dict()
+# learning policy
+lr_config = dict(policy='poly', power=0.9, min_lr=1e-4, by_epoch=False)


We may delete this file and modify the corresponding lr directly.

yamengxi · 2020-09-17T07:35:35Z

I have resolved all the problems, please review.

mmseg/models/segmentors/encoder_decoder.py

update hint Co-authored-by: Jerry Jiarui XU <xvjiarui0826@gmail.com>

yamengxi · 2020-09-18T03:29:41Z

Here is the update comment version.

xvjiarui · 2020-09-20T13:54:43Z

configs/deeplabv3/README.md

+### Pascal Context
+| Method | Backbone | Crop Size | Lr schd | Mem (GB) | Inf time (fps) | mIoU | mIoU(ms+flip) | download |
+|-----------|----------|-----------|--------:|----------|----------------|------:|--------------:|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| DeepLabV3 | R-101-D8 | 480x480 | 40000 | 9.2 | 7.09 | 46.55 | 47.81 | [model](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_480x480_80k_pascal_context/deeplabv3_r101-d8_480x480_40k_pascal_context_20200911_204118-1aa27336.pth) &#124; [log](https://openmmlab.oss-accelerate.aliyuncs.com/mmsegmentation/v0.5/deeplabv3/deeplabv3_r101-d8_480x480_80k_pascal_context/deeplabv3_r101-d8_480x480_40k_pascal_context-20200911_204118.log.json) |


These filenames are not the same as the paths provided. Please check.

Should be pascal_context- not pascal_context_.

Also, check the timestamp.

you means should be pascal_context_ not pascal_context-

yamengxi · 2020-09-22T04:37:36Z

I have fixed the model path mistakes.

hellock · 2020-09-22T14:22:14Z

Task linked: CU-4fp4de Pascal Context

* Add Pascal Context to mmsegmentation * Add benchmark result to Pascal Context * fix mmcv version * fix code syntax * fix code syntax again * Update mmseg/models/segmentors/encoder_decoder.py update hint Co-authored-by: Jerry Jiarui XU <xvjiarui0826@gmail.com> * update comment * fix pascal context model path * fix model path mistake again * fix model path mistake again * fix model path mistakes again Co-authored-by: Jerry Jiarui XU <xvjiarui0826@gmail.com>

* Update README.md line 23, 24 and 25: Remove "that" because "that" is unnecessary in these three sentences. line 33: Rewrite this sentence and make it more straightforward. line 34: This first sentence is incomplete. line 117: “focusses" -> "focuses" line 118: "continuous" -> "continuous" line 119: "consise" -> "concise" * Update README.md

Add Pascal Context to mmsegmentation

0999c55

xvjiarui reviewed Sep 15, 2020

View reviewed changes

tools/convert_datasets/pascal_context.py Show resolved Hide resolved

xvjiarui reviewed Sep 15, 2020

View reviewed changes

yamengxi added 4 commits September 17, 2020 14:26

Add benchmark result to Pascal Context

1c288b1

fix mmcv version

1efff53

fix code syntax

ed02391

fix code syntax again

f549690

xvjiarui reviewed Sep 17, 2020

View reviewed changes

mmseg/models/segmentors/encoder_decoder.py Outdated Show resolved Hide resolved

yamengxi and others added 3 commits September 18, 2020 11:16

Update mmseg/models/segmentors/encoder_decoder.py

4ec3d85

update hint Co-authored-by: Jerry Jiarui XU <xvjiarui0826@gmail.com>

update comment

496cd83

update comment

3835800

xvjiarui reviewed Sep 20, 2020

View reviewed changes

fix pascal context model path

cb4b6b6

yamengxi added 3 commits September 22, 2020 12:45

fix model path mistake again

dd04316

fix model path mistake again

8fed415

fix model path mistakes again

042901c

xvjiarui approved these changes Sep 22, 2020

View reviewed changes

Merge branch 'master' into Pascal_Context

1e73bf1

xvjiarui merged commit a2738fd into open-mmlab:master Sep 22, 2020

open-mmlab deleted a comment from OpenMMLab-Assistant-007 May 17, 2023

wjkim81 pushed a commit to wjkim81/mmsegmentation that referenced this pull request Dec 3, 2023

Support COCO-Wholebody (open-mmlab#133)

0d2568d

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add Pascal Context to mmsegmentation #133

Add Pascal Context to mmsegmentation #133

yamengxi commented Sep 15, 2020

CLAassistant commented Sep 15, 2020 •

edited

Loading

codecov bot commented Sep 15, 2020 •

edited

Loading

xvjiarui Sep 15, 2020

xvjiarui Sep 15, 2020

xvjiarui Sep 15, 2020

xvjiarui Sep 15, 2020

xvjiarui Sep 15, 2020

xvjiarui Sep 15, 2020

xvjiarui Sep 15, 2020

xvjiarui Sep 15, 2020

xvjiarui Sep 15, 2020

xvjiarui Sep 15, 2020

yamengxi commented Sep 17, 2020

yamengxi commented Sep 18, 2020

xvjiarui Sep 20, 2020

xvjiarui Sep 20, 2020

xvjiarui Sep 20, 2020

yamengxi Sep 22, 2020

yamengxi commented Sep 22, 2020

hellock commented Sep 22, 2020

		# assert h_crop <= h_img and w_crop <= w_img, (
		# 'crop size should not greater than image size')

		@@ -0,0 +1,15 @@
		GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=1 tools/slurm_train.sh mediaf pr_test configs/hrnet/fcn_hr48_480x480_40k_pascal_context.py --seed=0

Add Pascal Context to mmsegmentation #133

Add Pascal Context to mmsegmentation #133

Conversation

yamengxi commented Sep 15, 2020

CLAassistant commented Sep 15, 2020 • edited Loading

codecov bot commented Sep 15, 2020 • edited Loading

Codecov Report

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

yamengxi commented Sep 17, 2020

yamengxi commented Sep 18, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

yamengxi commented Sep 22, 2020

hellock commented Sep 22, 2020

CLAassistant commented Sep 15, 2020 •

edited

Loading

codecov bot commented Sep 15, 2020 •

edited

Loading