Return heatmaps and backbone features during inference #212

vsatyakumar · 2020-10-23T03:58:51Z

No description provided.

…g inference of top-down pose model

CLAassistant · 2020-10-23T03:58:56Z

All committers have signed the CLA.

innerlee · 2020-10-23T04:06:44Z

how does other repos return features? maybe use hooks? i'm not sure keep changing return signatures is a sustainable way

codecov · 2020-10-23T19:18:05Z

Codecov Report

Merging #212 into master will increase coverage by 0.31%.
The diff coverage is 41.62%.

@@            Coverage Diff             @@
##           master     #212      +/-   ##
==========================================
+ Coverage   82.90%   83.21%   +0.31%     
==========================================
  Files         106      108       +2     
  Lines        6211     6400     +189     
  Branches     1001     1030      +29     
==========================================
+ Hits         5149     5326     +177     
+ Misses        874      871       -3     
- Partials      188      203      +15

Flag	Coverage Δ
unittests	`83.21% <41.62%> (+0.31%)`	⬆️

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

Impacted Files	Coverage Δ
mmpose/datasets/datasets/__init__.py	`100.00% <ø> (ø)`
...datasets/datasets/top_down/topdown_mpii_dataset.py	`42.71% <ø> (ø)`
...sets/datasets/top_down/topdown_mpii_trb_dataset.py	`51.75% <ø> (ø)`
...datasets/datasets/top_down/topdown_coco_dataset.py	`88.44% <33.33%> (+24.71%)`	⬆️
...s/datasets/top_down/topdown_posetrack18_dataset.py	`34.07% <34.07%> (ø)`
mmpose/utils/hooks.py	`47.05% <47.05%> (ø)`
mmpose/apis/inference.py	`68.21% <52.94%> (-3.42%)`	⬇️
mmpose/models/detectors/top_down.py	`77.91% <92.30%> (-0.22%)`	⬇️
mmpose/datasets/datasets/top_down/__init__.py	`100.00% <100.00%> (ø)`
mmpose/datasets/pipelines/top_down_transform.py	`78.18% <0.00%> (-2.43%)`	⬇️
... and 8 more

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 3028b9f...2fffccb. Read the comment docs.

innerlee · 2020-10-24T15:55:57Z

there are linting errors. you can try to run pre-commit install on the root folder of the repo.
if pre-commit is not available, try pip install pre-commit

demo/top_down_img_demo.py

jin-s13 · 2020-10-25T14:36:01Z

Good!
One more thing. Since the return of forward_test is changed, the evaluate functions in the topdown datasets should also be modified accordingly, e.g.

mmpose/mmpose/datasets/datasets/top_down/topdown_coco_dataset.py

Line 322 in 3028b9f

for preds, boxes, image_path, _ in outputs:

to
for preds, boxes, image_path in outputs:

jin-s13 · 2020-10-25T14:53:39Z

Thanks!

And same issues here.

mmpose/mmpose/datasets/datasets/top_down/topdown_panoptic_dataset.py

Line 193 in 3028b9f

for preds, boxes, image_path, _ in outputs:

mmpose/mmpose/datasets/datasets/top_down/topdown_onehand10k_dataset.py

Line 180 in 3028b9f

for preds, boxes, image_path, _ in outputs:

mmpose/mmpose/datasets/datasets/top_down/topdown_mpii_trb_dataset.py

Line 179 in 3028b9f

for preds, boxes, image_path, _ in outputs:

mmpose/mmpose/datasets/datasets/top_down/topdown_freihand_dataset.py

Line 180 in 3028b9f

for preds, boxes, image_path, _ in outputs:

============================================

mmpose/mmpose/datasets/datasets/top_down/topdown_mpii_dataset.py

Line 153 in 3028b9f

preds = np.stack([kpts[0] for kpts, _, _, _ in outputs])

to

preds = np.stack([kpts[0] for kpts, _, _ in outputs])

innerlee · 2020-10-26T14:17:55Z

unit tests are failing, you can look at the error log here https://github.com/open-mmlab/mmpose/pull/212/checks?check_run_id=1305158220. you can also run pytest locally

vsatyakumar · 2020-10-26T15:15:39Z

Thanks, fixed build errors. Is this ready to merge?

mmpose/apis/inference.py

innerlee · 2020-10-30T07:08:54Z

Hi @vsatyakumar , we will release new version tomorrow. If the changes are quick enough, we can make this into this new version 😃

vsatyakumar · 2020-10-30T14:56:42Z

Hi @vsatyakumar , we will release new version tomorrow. If the changes are quick enough, we can make this into this new version 😃

Pushed fixes. Hope everything looks good. Thank you!

jin-s13

tests/data/smpl/test_joint_regressor.npy should not be added.

vsatyakumar · 2020-10-30T16:22:52Z

tests/data/smpl/test_joint_regressor.npy should not be added.

Yes, fixed. Please add test files to .gitignore

vsatyakumar · 2020-10-30T20:38:14Z

Thanks to @michalwols for feedback on improving hooks logic!

mmpose/utils/hooks.py

innerlee · 2020-10-31T04:11:47Z

@jin-s13 you may try this branch to see if it solves the memory issue

demo/top_down_img_demo.py

innerlee · 2020-10-31T14:18:27Z

Thanks!

* Added optional support to return heatmaps and backbone features during inference of top-down pose model * Added hooks for extracting layer outputs defined by name (and heatmap) * Fixed docstring * return heatmap condition fix * Docstring fixes * Docstring fixes * Docstring fixes * 'image' call fix * Fixed linting errors * fixed top down demo * fixed evaluate func * Fixed evaluate func * Fixed json loader in tests * Cleanup * try catch fix * try catch fix * Fixed model forward_test return signature to return heatmap * Deleted test file * Added .npy to ignore list * Optimized layer output hooks logic * Decoupled heatmap from output layer names. * Refactored mmpose.utils.hooks * Fixed mmpose.utils.hooks * Fixed output_layer_names default in top down demos Co-authored-by: OWAL Admin <janderoot@owal.io>

OpenMMLab-Assistant-007 · 2023-04-07T10:32:34Z

Hi!
@vsatyakumar
First of all, we want to express our gratitude for your significant PR in the OpenMMLab project. Your contribution is highly appreciated, and we are grateful for your efforts in helping improve this open-source project during your personal time. We believe that many developers will benefit from your PR.

We would also like to invite you to join our Special Interest Group (SIG) private channel on Discord, where you can share your experiences, ideas, and build connections with like-minded peers. To join the SIG channel, simply message moderator— OpenMMLab on Discord or briefly share your open-source contributions in the #introductions channel and we will assist you. Look forward to seeing you there! Join us ：https://discord.gg/UjgXkPWNqA

If you have WeChat account，welcome to join our community on WeChat. You can add our assistant ：openmmlabwx. Please add "mmsig + Github ID" as a remark when adding friends：）
Thank you again for your contribution❤

* Added optional support to return heatmaps and backbone features during inference of top-down pose model * Added hooks for extracting layer outputs defined by name (and heatmap) * Fixed docstring * return heatmap condition fix * Docstring fixes * Docstring fixes * Docstring fixes * 'image' call fix * Fixed linting errors * fixed top down demo * fixed evaluate func * Fixed evaluate func * Fixed json loader in tests * Cleanup * try catch fix * try catch fix * Fixed model forward_test return signature to return heatmap * Deleted test file * Added .npy to ignore list * Optimized layer output hooks logic * Decoupled heatmap from output layer names. * Refactored mmpose.utils.hooks * Fixed mmpose.utils.hooks * Fixed output_layer_names default in top down demos Co-authored-by: OWAL Admin <janderoot@owal.io>

Added optional support to return heatmaps and backbone features durin…

b777ddd

…g inference of top-down pose model

innerlee requested a review from jin-s13 October 23, 2020 04:05

Added hooks for extracting layer outputs defined by name (and heatmap)

698981a

vsatyakumar added 6 commits October 23, 2020 19:22

Fixed docstring

3aaf583

return heatmap condition fix

346b81c

Docstring fixes

8c3c6c0

Docstring fixes

972bc41

Docstring fixes

7166d64

'image' call fix

6cca49b

Fixed linting errors

e7b09db

jin-s13 reviewed Oct 25, 2020

View reviewed changes

demo/top_down_img_demo.py Outdated Show resolved Hide resolved

jin-s13 reviewed Oct 25, 2020

View reviewed changes

demo/top_down_img_demo.py Outdated Show resolved Hide resolved

fixed top down demo

c5d1452

fixed evaluate func

382ecf0

jin-s13 requested a review from innerlee October 25, 2020 14:54

Fixed evaluate func

94c6798

This comment has been minimized.

Sign in to view

Fixed json loader in tests

c66222c

innerlee reviewed Oct 26, 2020

View reviewed changes

mmpose/apis/inference.py Outdated Show resolved Hide resolved

vsatyakumar added 3 commits October 26, 2020 17:03

Cleanup

ad1209a

try catch fix

b9ca483

try catch fix

502754d

jin-s13 mentioned this pull request Oct 30, 2020

Roadmap of MMPose #9

Open

Fixed model forward_test return signature to return heatmap

c0933d6

jin-s13 reviewed Oct 30, 2020

View reviewed changes

Deleted test file

a50bc87

vsatyakumar added 2 commits October 30, 2020 16:34

Added .npy to ignore list

64f05cb

Optimized layer output hooks logic

e18ae69

Decoupled heatmap from output layer names.

4ca5391

innerlee reviewed Oct 31, 2020

View reviewed changes

mmpose/utils/hooks.py Outdated Show resolved Hide resolved

vsatyakumar added 2 commits October 31, 2020 03:52

Refactored mmpose.utils.hooks

383dad9

Fixed mmpose.utils.hooks

6941a72

innerlee reviewed Oct 31, 2020

View reviewed changes

demo/top_down_img_demo.py Outdated Show resolved Hide resolved

innerlee approved these changes Oct 31, 2020

View reviewed changes

Fixed output_layer_names default in top down demos

2fffccb

innerlee merged commit fb00a53 into open-mmlab:master Oct 31, 2020

innerlee mentioned this pull request Dec 2, 2020

How to get features from slowfastnet model open-mmlab/mmaction2#369

Closed

innerlee mentioned this pull request May 14, 2021

[Feature] Feature extraction open-mmlab/mmaction2#856

Merged

rollingman1 pushed a commit to rollingman1/mmpose that referenced this pull request Nov 5, 2021

[Enhance] Add cfg-options in argument (open-mmlab#212)

363fe9a

HAOCHENYE pushed a commit to HAOCHENYE/mmpose that referenced this pull request Jun 27, 2023

[Enhancement] Get local_rank in init_dist_mpi from env (open-mmlab#212)

a1adbff

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Return heatmaps and backbone features during inference #212

Return heatmaps and backbone features during inference #212

vsatyakumar commented Oct 23, 2020

CLAassistant commented Oct 23, 2020 •

edited

Loading

innerlee commented Oct 23, 2020

codecov bot commented Oct 23, 2020 •

edited

Loading

innerlee commented Oct 24, 2020

jin-s13 commented Oct 25, 2020

jin-s13 commented Oct 25, 2020

This comment has been minimized.

innerlee commented Oct 26, 2020

vsatyakumar commented Oct 26, 2020

innerlee commented Oct 30, 2020

vsatyakumar commented Oct 30, 2020

jin-s13 left a comment

vsatyakumar commented Oct 30, 2020

vsatyakumar commented Oct 30, 2020

innerlee commented Oct 31, 2020

innerlee commented Oct 31, 2020

OpenMMLab-Assistant-007 commented Apr 7, 2023

Return heatmaps and backbone features during inference #212

Return heatmaps and backbone features during inference #212

Conversation

vsatyakumar commented Oct 23, 2020

CLAassistant commented Oct 23, 2020 • edited Loading

innerlee commented Oct 23, 2020

codecov bot commented Oct 23, 2020 • edited Loading

Codecov Report

innerlee commented Oct 24, 2020

jin-s13 commented Oct 25, 2020

jin-s13 commented Oct 25, 2020

This comment has been minimized.

innerlee commented Oct 26, 2020

vsatyakumar commented Oct 26, 2020

innerlee commented Oct 30, 2020

vsatyakumar commented Oct 30, 2020

jin-s13 left a comment

Choose a reason for hiding this comment

vsatyakumar commented Oct 30, 2020

vsatyakumar commented Oct 30, 2020

innerlee commented Oct 31, 2020

innerlee commented Oct 31, 2020

OpenMMLab-Assistant-007 commented Apr 7, 2023

CLAassistant commented Oct 23, 2020 •

edited

Loading

codecov bot commented Oct 23, 2020 •

edited

Loading