How does the loss go if the network converges?

ZJULearning / ttfnet

Apache License 2.0

481 stars 96 forks source link

How does the loss go if the network converges? #18

Open Junwen-Huang opened 4 years ago

Junwen-Huang commented 4 years ago

I'm trying to train TFFNet on COCO2017, with resizing input to 256x256. Just for reference, could anyone give a range of the loss (or loss_heatmap, loss_wh) when it converges?

Libaishun commented 4 years ago

In my experiment, loss_wh can decrease to 0.21