[Enhance] Upgrage visualization to custom colors of different classes. by jbwang1997 · Pull Request #6716 · open-mmlab/mmdetection

jbwang1997 · 2021-12-07T15:02:36Z

Motivation

To custom colors for objects with different classes.

Modification

The main changes are in image.py
I also add a new argument named --show-colors in test.py to custom colors in the command line.

jbwang1997 · 2021-12-08T03:25:16Z

When learning visualization in mmdetection, I find the colors are still in the BGR format. It is a legacy of mmcv.visualization. Do you have a plan to update mmcv.visualization.

ZwwWayne · 2021-12-10T05:44:23Z

Hi @jbwang1997 ,
Thanks for your kind contribution. Could you fix the lint issue?

For now, we do not have a plan to upgrade mmcv.visualization. Do you mean that the arguments in the API implemented in mmdet still use the BGR format as input?

jbwang1997 · 2021-12-10T08:07:12Z

I tried to run pre-commit run --all-files, but got an ssl error.

Then I run isort mmdet/core/visualization/image.py, but also cannot pass the lint.

BIGWangYuDong · 2021-12-10T08:23:31Z

You can follow https://mmcv.readthedocs.io/en/latest/community/contributing.html#code-style to prepare pre-commit hook

jbwang1997 · 2021-12-10T11:04:33Z

You can follow https://mmcv.readthedocs.io/en/latest/community/contributing.html#code-style to prepare pre-commit hook

I have followed this ，but still reported same error.

hhaAndroid · 2021-12-13T02:45:19Z

tools/test.py

        default=0.3,
        help='score threshold (default: 0.3)')
+    parser.add_argument(
+        '--show-colors',


Please provide a simple example to illustrate the usage.

--show-colors can be a single color or a color file.

# single color input python tools/test.py \ configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py \ checkpoints/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth \ --show \ --show-colors green

# single color file python tools/test.py \ configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py \ checkpoints/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth \ --show \ --show-colors colors.txt

colors.txt

It is not a good way to load colors from txt file

BIGWangYuDong · 2021-12-15T11:15:31Z

Hi, we have discussed this PR today. Maybe you can follow MMSegmentation, an example: https://github.com/open-mmlab/mmsegmentation/blob/master/mmseg/datasets/coco_stuff.py. You can have a look at MMSeg codes, and use Dataset.PALETTE to set colors.

jbwang1997 · 2021-12-15T12:27:48Z

Ok, this is much better than directly setting colors in the command line. I will change the code as soon as possible.

BIGWangYuDong · 2021-12-15T12:34:21Z

Ok, this is much better than directly setting colors in the command line. I will change the code as soon as possible.

You can follow https://github.com/cocodataset/panopticapi/blob/master/panoptic_coco_categories.json to set labels color

Meanwhile, it is suggested to support random color, if labels are more than given colors.

jbwang1997 · 2021-12-15T12:38:54Z

@BIGWangYuDong I have a few questions. The color style in mmsegmentation is RGB. Do I need to change it into BGR to fit the mmcv.color_val or directly use the RGB colors and discard mmcv.color_val.

BIGWangYuDong · 2021-12-22T03:34:51Z

Please @AronLin have a look

demo/image_demo.py

mmdet/core/evaluation/class_names.py

mmdet/core/visualization/image.py

mmdet/datasets/custom.py

tools/train.py

AronLin · 2021-12-22T08:37:58Z

@BIGWangYuDong I have a few questions. The color style in mmsegmentation is RGB. Do I need to change it into BGR to fit the mmcv.color_val or directly use the RGB colors and discard mmcv.color_val.

We can directly use the RGB format palette.

jbwang1997 · 2021-12-29T14:30:13Z

Hello, I have added unit tests in the last commit. For updating the visualization image, could you tell me which image needs to update?

AronLin

LGTM

AronLin

Oh, we should modify the argument coments of imshow_det_bboxes. The color can be list now.

jbwang1997 · 2021-12-30T07:19:18Z

I have visualized some results of mask_rcnn_r50_fpn.

RangiLyu · 2022-01-07T02:22:06Z

mmdet/apis/inference.py

 def show_result_pyplot(model,
                       img,
                       result,
+                       palette=None,
                       score_thr=0.3,
                       title='result',
                       wait_time=0):


Avoid break using positional arguments.

Suggested change

def show_result_pyplot(model,

img,

result,

palette=None,

score_thr=0.3,

title='result',

wait_time=0):

def show_result_pyplot(model,

img,

result,

score_thr=0.3,

title='result',

wait_time=0,

palette=None,):

Thank you for your suggestion.

mmdet/core/visualization/image.py

RangiLyu · 2022-01-17T03:24:27Z

Panoptic visualization is based on this PR. We should merge it asap. Kindly ping @ZwwWayne .

open-mmlab#6716) * Update visualization * Fix a help text mistake * short function docs * Fix for iosrt * Back to master * Check by pre-commit * Update dataset_warppers palette * Custom colors for visualizaton * Update visualization in image_demo.py * align palette with panopticapi * delete palette from checkpoint meta * Use tuple to represent color * Update visualization * Modify image_demo.py * Check test error * Check test error * Fix error for empty case * Delete redundant functions * Check visualization * Add default case * Fix bug * Fix bugs * Add unit tests * Change coments * Avoid break using positional arguments. * Add copyright * Modify comments * fix lint Co-authored-by: Guangchen Lin <347630870@qq.com>

#6716) * Update visualization * Fix a help text mistake * short function docs * Fix for iosrt * Back to master * Check by pre-commit * Update dataset_warppers palette * Custom colors for visualizaton * Update visualization in image_demo.py * align palette with panopticapi * delete palette from checkpoint meta * Use tuple to represent color * Update visualization * Modify image_demo.py * Check test error * Check test error * Fix error for empty case * Delete redundant functions * Check visualization * Add default case * Fix bug * Fix bugs * Add unit tests * Change coments * Avoid break using positional arguments. * Add copyright * Modify comments * fix lint Co-authored-by: Guangchen Lin <347630870@qq.com>

open-mmlab#6716) * Update visualization * Fix a help text mistake * short function docs * Fix for iosrt * Back to master * Check by pre-commit * Update dataset_warppers palette * Custom colors for visualizaton * Update visualization in image_demo.py * align palette with panopticapi * delete palette from checkpoint meta * Use tuple to represent color * Update visualization * Modify image_demo.py * Check test error * Check test error * Fix error for empty case * Delete redundant functions * Check visualization * Add default case * Fix bug * Fix bugs * Add unit tests * Change coments * Avoid break using positional arguments. * Add copyright * Modify comments * fix lint Co-authored-by: Guangchen Lin <347630870@qq.com>

jbwang1997 added 3 commits December 7, 2021 22:44

Update visualization

3f362c8

Fix a help text mistake

569b0e2

short function docs

bdb6c87

ZwwWayne requested review from BIGWangYuDong, hhaAndroid and jshilong December 8, 2021 03:03

Fix for iosrt

0dc5c05

hhaAndroid reviewed Dec 13, 2021

View reviewed changes

BIGWangYuDong changed the title ~~Upgrage visualization to custom colors of different classes.~~ [Enhance] Upgrage visualization to custom colors of different classes. Dec 15, 2021

jbwang1997 added 5 commits December 18, 2021 20:51

Back to master

bea1a9f

Check by pre-commit

b88346b

Update dataset_warppers palette

4bb93f4

Custom colors for visualizaton

0128098

Update visualization in image_demo.py

e111790

jbwang1997 requested a review from hhaAndroid December 19, 2021 14:43

BIGWangYuDong requested a review from AronLin December 22, 2021 03:34

AronLin reviewed Dec 22, 2021

View reviewed changes

jbwang1997 added 2 commits December 25, 2021 17:11

align palette with panopticapi

374dfc8

delete palette from checkpoint meta

020a234

Add unit tests

f5e178e

AronLin self-requested a review December 30, 2021 05:32

AronLin approved these changes Dec 30, 2021

View reviewed changes

AronLin reviewed Dec 30, 2021

View reviewed changes

Change coments

0a847bd

AronLin requested a review from RangiLyu January 5, 2022 03:21

RangiLyu reviewed Jan 7, 2022

View reviewed changes

jbwang1997 added 2 commits January 7, 2022 11:50

Avoid break using positional arguments.

f1774e5

Add copyright

ef0fd6a

RangiLyu mentioned this pull request Jan 10, 2022

Color setting for bounding box #6980

Closed

ZwwWayne reviewed Jan 13, 2022

View reviewed changes

mmdet/core/visualization/image.py Outdated Show resolved Hide resolved

AronLin added 2 commits January 13, 2022 20:28

Modify comments

ace2946

fix lint

d247486

RangiLyu approved these changes Jan 13, 2022

View reviewed changes

RangiLyu changed the base branch from master to dev January 17, 2022 03:22

ZwwWayne approved these changes Jan 17, 2022

View reviewed changes

ZwwWayne merged commit b6f21e3 into open-mmlab:dev Jan 17, 2022

gaotongxiao mentioned this pull request Jan 27, 2022

[Fix] Relax the requirement of PALETTE in dataset wrappers #7085

Merged

AronLin mentioned this pull request Feb 16, 2022

How to specify a palette instead of a single color in mask_color in model.show_result #7161

Closed

zytx121 mentioned this pull request Feb 19, 2022

--show-dir error open-mmlab/mmrotate#9

Closed

jbwang1997 mentioned this pull request Feb 20, 2022

[Fix] fix visualization bugs open-mmlab/mmrotate#12

Merged

zytx121 mentioned this pull request Mar 11, 2022

Visualize boxes of diff classes as diff colors open-mmlab/mmrotate#121

Closed

liminyu-web mentioned this pull request Mar 17, 2022

TypeError: show_result() got an unexpected keyword argument 'mask_color' open-mmlab/mmrotate#144

Closed

liuyanyi mentioned this pull request Apr 5, 2022

[Enhance] Refactor imshow_det_rbboxes open-mmlab/mmrotate#187

Merged

Conversation

jbwang1997 commented Dec 7, 2021

Motivation

Modification

Uh oh!

jbwang1997 commented Dec 8, 2021

Uh oh!

ZwwWayne commented Dec 10, 2021

Uh oh!

jbwang1997 commented Dec 10, 2021

Uh oh!

BIGWangYuDong commented Dec 10, 2021

Uh oh!

jbwang1997 commented Dec 10, 2021

Uh oh!

hhaAndroid Dec 13, 2021

Choose a reason for hiding this comment

Uh oh!

jbwang1997 Dec 13, 2021

Choose a reason for hiding this comment

Uh oh!

BIGWangYuDong Dec 15, 2021

Choose a reason for hiding this comment

Uh oh!

BIGWangYuDong commented Dec 15, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jbwang1997 commented Dec 15, 2021

Uh oh!

BIGWangYuDong commented Dec 15, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jbwang1997 commented Dec 15, 2021

Uh oh!

BIGWangYuDong commented Dec 22, 2021

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

AronLin commented Dec 22, 2021

Uh oh!

jbwang1997 commented Dec 29, 2021

Uh oh!

AronLin left a comment

Choose a reason for hiding this comment

Uh oh!

AronLin left a comment

Choose a reason for hiding this comment

Uh oh!

jbwang1997 commented Dec 30, 2021

Uh oh!

RangiLyu Jan 7, 2022

Choose a reason for hiding this comment

Uh oh!

jbwang1997 Jan 7, 2022

Choose a reason for hiding this comment

Uh oh!

Uh oh!

RangiLyu commented Jan 17, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

BIGWangYuDong commented Dec 15, 2021 •

edited

Loading

BIGWangYuDong commented Dec 15, 2021 •

edited

Loading