Skip to content
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.

Sign up for GitHub

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

Jump to bottom

Code for Paper "Bridging the Gap Between Anchor-based and Anchor-free… #1872

Merged
merged 7 commits into from
Jan 21, 2020
Merged

Code for Paper "Bridging the Gap Between Anchor-based and Anchor-free… #1872

merged 7 commits into from
Jan 21, 2020

Conversation

hust-kevin
Copy link
Contributor

No description provided.

Code for Paper "Bridging the Gap Between Anchor-based and Anchor-free…
c651660 
… Detection via Adaptive Training Sample Selection"
@hellock hellock requested a review from yhcao6 December 26, 2019 07:06
@hust-kevin
Copy link
Contributor Author

we re-implement the atss, more detail refer to office repo and paper

Re-implement Note

the core of this paper is Adaptive Training Sample Selection, while the final model (coco ap 39.2) actually consists of the following parts.

ATSS (final model) = RetinaNet(#A=1) + RetinaNet tricks + Imprs on FCOS + atss

  • RetinaNet(#A=1) means only one square anchor box per location in RetinaNet
  • RetinaNet tricks is some trick apply in RetinaNet which boost ap from 35.6 to 36.3, this trick include use std [0.1, 0.1, 0.2, 0.2] and classification loss normalization, in the paper, this is used as the baseline for RetinaNet. more detail refer to maskrcnn-benckmark. (PS: we don't do experiment on mmdection whether it can get the same boost)
  • Imprs on FCOS is some trick used in new version of fcos to further impore the result, which include use GN, GIOU Loss, Centerness etc
  • atss is the core part of this paper, which is a bbox assigner mechanism to replace the origin max-iou assigner, more detail refer to paper and code in core/bbox/assigners/atss_assigner.py

yhcao6
yhcao6 suggested changes Jan 7, 2020
View reviewed changes
Copy link
Collaborator

@yhcao6 yhcao6 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for your hard work! I make some suggestions on some part of the code. Feel free to point out if I make some mistakes.

configs/atss/README.md Outdated Show resolved Hide resolved
mmdet/core/anchor/atss_target.py Outdated Show resolved Hide resolved
mmdet/core/anchor/atss_target.py Outdated Show resolved Hide resolved
mmdet/core/bbox/assigners/atss_assigner.py Show resolved Hide resolved
mmdet/core/bbox/assigners/atss_assigner.py Outdated Show resolved Hide resolved
mmdet/models/anchor_heads/atss_head.py Outdated Show resolved Hide resolved
mmdet/models/anchor_heads/atss_head.py Outdated Show resolved Hide resolved
mmdet/models/anchor_heads/atss_head.py Show resolved Hide resolved
mmdet/models/losses/iou_loss.py Show resolved Hide resolved
mmdet/models/losses/iou_loss.py Outdated Show resolved Hide resolved
@hust-kevin
Copy link
Contributor Author

@yhcao6 we re-construct and re-train the model, now ap is 39.4

@yhcao6
Copy link
Collaborator

yhcao6 commented Jan 9, 2020

@yhcao6 we re-construct and re-train the model, now ap is 39.4

It is awesome! I would like to add on more that in my experiment, +1 in GIoU loss also boost the performance.
I am benchmarking the model on 8gpus since almost all models are trained on 8gpus. I will inform you if there is new progress.
Thanks for you quick reply.

@yhcao6 yhcao6 self-requested a review January 14, 2020 02:56
@hellock hellock merged commit 1ebb6cb into open-mmlab:master Jan 21, 2020
mattdawkins added a commit to VIAME/mmdetection that referenced this pull request Mar 13, 2020
@mattdawkins
Merge remote-tracking branch 'origin/viame/master' into viame/master-old
3eaa2a7 
* origin/viame/master: (28 commits)
  Fix FPN upscale
  Extra compiler args
  VIAME-specific build parameters
  Bump version to 1.0.0 (open-mmlab#2029)
  Fix the incompatibility of the latest numpy and pycocotools (open-mmlab#2024)
  format configs with yapf (open-mmlab#2023)
  options for FCNMaskHead during testtime (open-mmlab#2013)
  Enhance AssignResult and SamplingResult (open-mmlab#1995)
  Fix typo activatation -> activation (open-mmlab#2007)
  Reorganize requirements, make albumentations optional (open-mmlab#1969)
  Encapsulate DCN into a ConvModule & Conv_layers (open-mmlab#1894)
  Code for Paper "Bridging the Gap Between Anchor-based and Anchor-free… (open-mmlab#1872)
  Non color images (open-mmlab#1976)
  Fix albu mask format bug (open-mmlab#1818)
  Fix CI by limiting the version of torchvision (open-mmlab#2005)
  Add ability to overwite existing module in Registry (open-mmlab#1982)
  bug for distributed training (open-mmlab#1985)
  Update Libra RetinaNet config with the latest code (open-mmlab#1975)
  Fix issue in refine_bboxes and add doctest (open-mmlab#1962)
  add link to official repo (open-mmlab#1971)
  ...
ioir123ju pushed a commit to ioir123ju/mmdetection that referenced this pull request Mar 30, 2020
@hust-kevin @yhcao6 @ioir123ju
Code for Paper "Bridging the Gap Between Anchor-based and Anchor-free… (
06ec29e 
open-mmlab#1872)

* Code for Paper "Bridging the Gap Between Anchor-based and Anchor-free Detection via Adaptive Training Sample Selection"

* fix format

* changed the code according to the Changes requested

* update benchmark

* minor refactoring

* minor fix

* update model_zoo and support models

Co-authored-by: Cao Yuhang <yhcao6@gmail.com>
mike112223 pushed a commit to mike112223/mmdetection that referenced this pull request Aug 25, 2020
@hust-kevin @yhcao6 @hellock
Code for Paper "Bridging the Gap Between Anchor-based and Anchor-free… (
33536d6 
open-mmlab#1872)

* Code for Paper "Bridging the Gap Between Anchor-based and Anchor-free Detection via Adaptive Training Sample Selection"

* fix format

* changed the code according to the Changes requested

* update benchmark

* minor refactoring

* minor fix

* update model_zoo and support models

Co-authored-by: Cao Yuhang <yhcao6@gmail.com>
@OpenMMLab-Assistant-004
Copy link

Hi @hust-kevin ！First of all, we want to express our gratitude for your significant PR in the mmdetection 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/raweFPmdzG

If you have WeChat，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❤

1 similar comment
@OpenMMLab-Assistant-004
Copy link

Hi @hust-kevin ！First of all, we want to express our gratitude for your significant PR in the mmdetection 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/raweFPmdzG

If you have WeChat，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❤

FANGAreNotGnu pushed a commit to FANGAreNotGnu/mmdetection that referenced this pull request Oct 23, 2023
@sxjscience
Port AutoMMPredictor to Tabular + Add kaggle house competition exampl…
abca097 
…e + Improve example readmes (open-mmlab#1872)

* update

* fix

* fix

* Update run_all.sh

* Update example_tabular.py

* fix

* update

* Update automm_model.py

* Update automm_model.py

* update

* fix

* Update automm_model.py
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@yhcao6 yhcao6 yhcao6 approved these changes

@hellock hellock hellock approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@hust-kevin @yhcao6 @OpenMMLab-Assistant-004 @hellock