-
Notifications
You must be signed in to change notification settings - Fork 333
Anchor free models #1
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
Comments
Very naive questions here:
I would be very happy to discuss further on this subject, and maybe try to participate on this project, as I m currently trying to implement a (very humble) anchor based "old school" vision control library. |
Anchor free models are probably not in our immediate roadmaps yet. @fchollet WDYT? |
@tanzhenyu At least it would have a bit of a way to diversify keras-cv compared to the model garden since almost all official anchor free papers implementations are in pytorch. EDIT: |
True -- but we'll focus on things not TF2/Keras implemented yet, such as semantic segmentation, DeepLab, etc. If this is in the long run outperforming anchor-based models, (as well as DETR I guess), then we'd need to consider having re-useable layers for them. So "long run" is the vague word here, would it make more sense to put in addons? |
If you are talking about official models is true. But there are many third_party very good impl related to your list on Github. |
Yes -- and I'd like to consolidate the efforts. If that's already very good impl we should ask the author to migrate directly to this repo (or Keras-applications if it's backbone) |
@tanzhenyu nevermind anchor-free just appeared in the new model garden object detection API (Centernet) https://github.com/tensorflow/models/tree/master/research/object_detection#tensorflow-2-support |
IMHO less reusable when we have utils in models repos https://github.com/tensorflow/models/blob/master/research/object_detection/utils/target_assigner_utils.py#L44 |
* add fmix init commit * fix issues try #1 * fix issues try #2 * change random generator to normal from uniform to support negative values * add support for variable channels attempt 1 * update fftfreq description * add changes * add support for variable label shapes * add changes * fix errors * remove newline * change name to FourierMix attempt 1 * fixed lint issue * remove edge softening * Final pass Co-authored-by: Luke Wood <lukewoodcs@gmail.com>
* add fmix init commit * fix issues try #1 * fix issues try #2 * change random generator to normal from uniform to support negative values * add support for variable channels attempt 1 * update fftfreq description * add changes * add support for variable label shapes * add changes * fix errors * remove newline * change name to FourierMix attempt 1 * fixed lint issue * remove edge softening * Final pass Co-authored-by: Luke Wood <lukewoodcs@gmail.com>
* add fmix init commit * fix issues try keras-team#1 * fix issues try keras-team#2 * change random generator to normal from uniform to support negative values * add support for variable channels attempt 1 * update fftfreq description * add changes * add support for variable label shapes * add changes * fix errors * remove newline * change name to FourierMix attempt 1 * fixed lint issue * remove edge softening * Final pass Co-authored-by: Luke Wood <lukewoodcs@gmail.com>
…d update iou losses (#1296) * first attempt at introducing YoloX * formatted and fixed bugs * cast fix #1 * cast fix #2 * cast fix #3 * cast fix #4 * adding ensure shape for support * reverting and removing ensure_shape * fixed another bug * updated train.py * updated docs, tests and added support for loss strings * first attempt at introducing YoloX * formatted and fixed bugs * adding ensure shape for support * reverting and removing ensure_shape * reformatted by black * fixed a linting issue * finally rebased atop the recent changes * finally rebased atop the new changes * fixed linting issues * reverted rebasing issues with iou loss * fixing rebased errors part 2 * fixed more linting issues * TPU testing changes * linting fixes * updated with implementation details from paper * updated based on review comments and api changes * first attempt at introducing YoloX * updated docs, tests and added support for loss strings * fixed linting issues * reverted rebasing issues with iou loss * review comments * removed examples * linting fix * fixed rebasing error * updated no_reduction warning * review comments * revert version and linting fixes
* Add numpy ops (initial batch) and some config * Add unit test * fix call * Revert "fix call" This reverts commit 6748ad183029ff4b97317b77ceed8661916bb9a0. * full unit test coverage * fix setup.py
…d update iou losses (keras-team#1296) * first attempt at introducing YoloX * formatted and fixed bugs * cast fix keras-team#1 * cast fix keras-team#2 * cast fix keras-team#3 * cast fix keras-team#4 * adding ensure shape for support * reverting and removing ensure_shape * fixed another bug * updated train.py * updated docs, tests and added support for loss strings * first attempt at introducing YoloX * formatted and fixed bugs * adding ensure shape for support * reverting and removing ensure_shape * reformatted by black * fixed a linting issue * finally rebased atop the recent changes * finally rebased atop the new changes * fixed linting issues * reverted rebasing issues with iou loss * fixing rebased errors part 2 * fixed more linting issues * TPU testing changes * linting fixes * updated with implementation details from paper * updated based on review comments and api changes * first attempt at introducing YoloX * updated docs, tests and added support for loss strings * fixed linting issues * reverted rebasing issues with iou loss * review comments * removed examples * linting fix * fixed rebasing error * updated no_reduction warning * review comments * revert version and linting fixes
Please evaluate to extract common components/utils that could support Anchor free models:
tensorflow/hub#424
See also:
google-ai-edge/mediapipe#495 (comment)
google-coral/edgetpu#51
The text was updated successfully, but these errors were encountered: