-
Notifications
You must be signed in to change notification settings - Fork 9.4k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Refactor] Add simple_test to dense heads #5061
Conversation
Codecov Report
@@ Coverage Diff @@
## master #5061 +/- ##
==========================================
- Coverage 65.33% 65.28% -0.05%
==========================================
Files 276 275 -1
Lines 21204 21215 +11
Branches 3522 3524 +2
==========================================
- Hits 13854 13851 -3
- Misses 6601 6617 +16
+ Partials 749 747 -2
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
Thank you for your contribution, I think it is very good. Can you resolve the conflict? |
…to clean_simple_test
The outer list corresponds to each image. The inner list | ||
corresponds to each class. | ||
""" | ||
return self.simple_test_bboxes( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Seems you add this arguments postprocess
to get raw results of boxes in yolact
, but I believe original design keep bbbox2results
is single-stage would a better design and can do the same thing, Is there other reasons?
In addition, I think keeping this operation in the SingleStageDetector
instead of simple_test_bboxes
would be more consistent with the two-tage models.
just like
bbox_results = [ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I added the argument postprocess
to simplify code for onnx.
mmdetection/mmdet/models/detectors/single_stage.py
Lines 117 to 118 in 7578b6c
# skip post-processing when exporting to ONNX | |
postprocess = False |
The original reason disappeared after #5205
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The overall design looks good to me, and I would help to move bbbox2results
to simple_test
of single_stage.py.
It would be merged soon, Thanks for your contribution.
…arable performance (open-mmlab#5136) * Supports for exporting CornerNet to ONNX with dynamic shapes and comparable performance * add docs for exporting cornernet, and simplify code * fix doc * format doc * fix docstring
* Update fcn_mask_head.py * Update fcn_mask_head.py * Update fcn_mask_head.py * Resolve format issues Co-authored-by: Wenwei Zhang <40779233+ZwwWayne@users.noreply.github.com>
* evaluate trt models * update version of onnx * update maskrcnn results * add backend argument * update fcos results * update * fix bug * update doc
* update changelog * update changelog * Improvements * Improvements * update for v2.13.0 Co-authored-by: hhaAndroid <1286304229@qq.com>
Merged through #5264. So this PR is closed. |
This PR improves consistency of implementation (single-stage vs. two-stage, normal dense heads vs. RPN heads, simple_test vs. aug_test).
Related discussion: #3638 (review)