Why use SPPCSPC instead of SPPFCSPC？

WongKinYiu / yolov7

Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

GNU General Public License v3.0

13.22k stars 4.18k forks source link

Why use SPPCSPC instead of SPPFCSPC？ #658

Open WangQvQ opened 2 years ago

WangQvQ commented 2 years ago

Thank you for your work, but I have a question. yolov5's SPPF is much faster than SPP. Why not try to replace SPPCSPC with SPPFCSPC? Hope to receive your reply☺️ SPPCSPC(1)

WongKinYiu commented 2 years ago

Max pooling uses very few computation, if you programming well, above one could run three max pool layers in parallel, while below one must process three max pool layers sequentially.

By the way, you could replace SPPCSPC by SPPFCSPC at inference time if your hardware is friendly to SPPFCSPC.

WangQvQ commented 2 years ago

Thank you for your reply, I get it Have a nice day!

WangQvQ commented 1 year ago

Amazing! Meituan just used this structure in YOLOv6 3.0, do I need to contribute this code in YOLOv7？

WongKinYiu commented 1 year ago

Yes, you could contribute code.

By the way, I think YOLOv6 v3.0 should give some credit to DATE about AAT while their proposed AAT share same dual assignment idea and architecture.