search

Media-Smart / vedadet

A single stage object detection toolbox based on PyTorch
Apache License 2.0
498 stars 128 forks source link

Question on TinaFace training setting? #58

Closed huangh12 closed 3 years ago

huangh12 commented 3 years ago

Hi @hxcai , thanks for sharing the impressive work. I have read the TinaFace paper and been puzzled by the training setting in the paper. The lr schedule and lr rate seems very complex and differs a lot from the official RetinaNet. Is this training setting a careful tunning result of the experiments or being inspired by other work. Could you reveal the motivation behind that? Thank you!

图片

hxcai commented 3 years ago

@huangh12 Hi，we use the restart learning scheduler to conduct experiments for convenience and do not tune carefully.

huangh12 commented 3 years ago

Thank you.