search

ultralytics / yolov5

YOLOv5 🚀 in PyTorch > ONNX > CoreML > TFLite
https://docs.ultralytics.com
GNU Affero General Public License v3.0
50.6k stars 16.32k forks source link

Will support Recursive-FPN？ #363

Closed zhongqiu1245 closed 4 years ago

zhongqiu1245 commented 4 years ago

Hey, thank you for sharing your amazing job! Will you increase support for Recursive-FPN in the future?

glenn-jocher commented 4 years ago

@zhongqiu1245 I'm not familiar with it. Can you provide more details?

zhongqiu1245 commented 4 years ago

https://arxiv.org/abs/2006.02334 https://github.com/joe-siyuan-qiao/DetectoRS 《DetectoRS: Detecting Objects with Recursive Feature Pyramid and Switchable Atrous Convolution》 "On COCO test-dev, DetectoRS achieves state-of-theart 54.7% box AP for object detection, 47.1% mask AP for instance segmentation, and 49.6% PQ for panoptic segmentation."

Recursive-FPN is the output of the fusion of the traditional FPN, input again to Backbone(a second cycle), and finally sent to the Head. I think that may be a new Bag-of-Specials and may be helpful to improve the AP of yolov5

github-actions[bot] commented 4 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.