DataXujing / YOLOv9

:fire: YOLOv9 paper解析,训练自己的数据集,TensorRT端到端部署, NCNN安卓手机部署
GNU General Public License v3.0
57 stars 11 forks source link

train.py 和 train_dual.py 的区别是什么? #1

Closed zhazhahao closed 6 months ago

zhazhahao commented 6 months ago

train.py 和 train_dual.py 的区别是什么?

DataXujing commented 6 months ago

可以参考这个解释:https://github.com/WongKinYiu/yolov9/issues/1 train_dual.py是有一个辅助分支和一个主分支的训练代码。

train.py: no aux branch + one main branch
train_dual: one aux branch + one main branch.
triple_branch: two aux branches + one main branch

We provide these codes to show how to extend aux branches for training. Everyone could follow the rule to make related researches. For example, we also extend train_multiple to implement training strategy which is similar to [Co-DETR](https://openaccess.thecvf.com/content/ICCV2023/html/Zong_DETRs_with_Collaborative_Hybrid_Assignments_Training_ICCV_2023_paper.html).
zhazhahao commented 6 months ago

谢谢博主