wqz960

@1adrianb. I saw many papers splits the aflw2000 to 3parts (0,30) (30,60) (60,90) . Is there a ruler to get the 3 parts? Thank you

Thank you for your kind help! Can you provide me some network structures for 256 size, the vunet is open-source for tensorflow, it is too difficult for me to re-implement...

@testingshanu @ElPatou have you train the network again? can you get the same training nme with the author's?

another question: When I use two gpus, how to change the learning rate? lr*2?

@yangyangkiki Is there some tricks I forgot to add? I followed your Readme, but the training on 300W dataset was not the same with the log you provided. Finally, the...

@CrossEntropy I train the model, but it does not convergence, have you met this problem?

@CrossEntropy I did not clean the data, use all the pictures(about 30K) for training. the lowest the losses of yaw, pitch and roll are about +-1.5, and I evaluate the...

hello can you tell me the lowest losses for yaw pitch and roll? Thanks @natanielruiz

@dfzsgjshzfj Using the original datapreprocessing code, the loss will drop to +-1.5, but the result on AFLW2000 is bad, I have given it up, turning to another repo.

@rosebbb emmmm. sorry to bother you! can you provide me the script for transforming the ground true landmarks to the boundary maps instead of a test model? Thank you!