deep-head-pose-lite
deep-head-pose-lite copied to clipboard
stevenyangyj
Train problem
Dear author, tnanks for sharing so wonderful work. I want to do a head pose task in wild world. The hopenet runs too slow and I found your work. I test it and it runs so fast, but the performance is poor. So I want to train my own model based on your work. I can't find train details in your github page. Can you provide some help on training? How to train model?Thanks!
Dear author, tnanks for sharing so wonderful work. I want to do a head pose task in wild world. The hopenet runs too slow and I found your work. I test it and it runs so fast, but the performance is poor. So I want to train my own model based on your work. I can't find train details in your github page. Can you provide some help on training? How to train model?Thanks!
Please refer to the training code of original project, and sorry, I cannot give you specific advice...
@OverEuro, thanks for your great work. It is really helpful for me. in training code of original project, the default alpha is 0.001, defalut epoch is 5, and learing rate is 0.001. How about yours , can you share your training code?
@OverEuro, thanks for your great work. It is really helpful for me. in training code of original project, the default alpha is 0.001, defalut epoch is 5, and learing rate is 0.001. How about yours , can you share your training code?
check this issue and I feel like SGD optimizer with warm-up might be a better choice.
Hi @stevenyangyj,
I was trying to train your model again to check if the accuracies improve. However, I am getting error in optimizer line which I suppose is because of the missing layers or different architecture of the shufflenet model used. It will be very helpful if you can provide your setting for the training of the model to reproduce the results.
AttributeError: 'HopeNetLite' object has no attribute 'bn1'
