search

Stinky-Tofu / Stronger-yolo

🔥Improve yolo with latest paper
MIT License
3 stars 0 forks source link

mixup weight #54

Open WangDeyu opened 5 years ago

WangDeyu commented 5 years ago

Why do you use loss (include GIOU_loss, conf_loss, prob_loss ) to mutiply label_mixw? From the paper, label_mixw is just responsible for the prob_loss.

WangDeyu commented 5 years ago

And in https://arxiv.org/pdf/1902.04103.pdf, it doesn't mention the mixup_weight for label, maybe we can use the label combination of two image directly？

Stinky-Tofu commented 5 years ago

Please read the paper carefully https://arxiv.org/pdf/1902.04103.pdf @WangDeyu

chestnutLeeeee commented 5 years ago

老哥我跟你的想法一样，loss前不用乘mixw

WangDeyu commented 5 years ago

老哥我跟你的想法一样，loss前不用乘mixw

哥们，貌似是要乘的，后来我又看了遍https://arxiv.org/pdf/1902.04103.pdf, 你仔细看一下Table 1里的weighted loss说明