laodar / cnn_head_pose_estimator Goto Github PK

How can i train the model using my own dataset, can you please provide the training code and explanation how to train

Hi, I am wondering whether there is a reference for the extract_image_chips method?

Thanks a lot,
Tobias

Hi, in an earlier communication, you said the performance in real scene is not very good. Do you get a chance to retrain your model with real scene images? And can you share the training code?
Thanks a lot.

Hi, can you explain a little bit the exact definition of the two numbers in the pitch-yaw angles?
Thanks a lot.

RuntimeError: simple_bind error. Arguments:
fullyconnected1_bias: (2,)
convolution3_bias: (64,)
fullyconnected1_weight: (2, 128)
convolution1_weight: (32, 32, 3, 3)
fullyconnected0_bias: (128,)
convolution1_bias: (32,)
convolution3_weight: (64, 64, 3, 3)
convolution2_weight: (64, 32, 3, 3)
convolution0_weight: (32, 3, 3, 3)
fullyconnected0_weight: (128, 576)
convolution0_bias: (32,)
convolution2_bias: (64,)
data: (1, 3, 64, 64)
Error in operator ls: [16:20:16] c:\jenkins\workspace\mxnet\mxnet\include\mxnet\tuple.h:202: Check failed: i >= 0 && i < ndim(): index = 0 must be in range [0, -1)

the kinect data have not provide the face box ,eg(x,y,w,h);so you use oher detection method get the face box? like opencv or mtcnn?