search

yfeng95 / PRNet

Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network (ECCV 2018)
http://openaccess.thecvf.com/content_ECCV_2018/papers/Yao_Feng_Joint_3D_Face_ECCV_2018_paper.pdf
MIT License
4.95k stars 947 forks source link

Training PRN model #93

Open rafaelbou opened 5 years ago

rafaelbou commented 5 years ago

Hi everyone, I began to train the model and I have some questions:

  1. Dataset: I downloaded 300W_LP dataset and it contains ~200K images while the author report about ~60K images at the same dataset. Someone know what is the difference between these versions?

  2. Overfitting: While training the dataset I'm getting in to overfit: image someone manage same issue too?

  3. About the loss function: Do you add L2 regularization to the loss function? Do you use the uv_weight_mask.png as it is? Do you multiply it with uv_face_mask.png to ignore the neck region? Do you use L1 or L2 for the loss calculation (uv' - uv)?

  4. Evaluation: How do you evaluate your model? meanwhile I'm using only the loss function and visualize the output uv maps.

Thanks to the answers, Rafael.

Originally posted by @rafaelbou in https://github.com/YadiraF/PRNet/issues/46#issuecomment-446916937

qiannvMao commented 5 years ago

@rafaelbou Hi,as a newer to this, I failed to achieve the training code, could you please provide training code to me? I hope I can get your help, thank you very much. My email: mqn20170910@163.com

lxzkeyan commented 4 years ago

@rafaelbou Can you share your training code?Looking forward to your reply.Thank you again.Email address: lxz_nxu@163.com

xingxingsuo33 commented 4 years ago

@rafaelbou Can you share your training code?Looking forward to your reply.Thank you again.Email address: 549012844@qq.com

yanghao960502 commented 3 years ago

可以分享一下你的训练代码吗？期待你的回复。再次感谢。邮箱 1363627028@qq.com