mpatacchiola / deepgaze

Computer Vision library for human-computer interaction. It implements Head Pose and Gaze Direction Estimation Using Convolutional Neural Networks, Skin Detection through Backprojection, Motion Detection and Tracking, Saliency Map.
MIT License
1.79k stars 478 forks source link

[ex_aflw_parser] #61

Open speculaas opened 6 years ago

speculaas commented 6 years ago

Dear mpatacchiola,

Thanks for your help!

https://github.com/mpatacchiola/deepgaze/blob/master/examples/ex_dnn_head_pose_estimation_training/ex_aflw_parser.py

In the above file:

Separating the label from the dataset

dataset = data[:,0:4096] label = data[:,4096:4099]

I think length of the image in 1D should be 64*64 = 4096? It's 1600 in the current commit and I got error when I try to run ex_dnn_head_pose_estimation_training.py :

Training set (19507, 1600) (19507, 3) Validation set (2438, 1600) (2438, 3) Test set (2439, 1600) (2439, 3) Starting Graph creation... Traceback (most recent call last): File "ex_dnn_head_pose_estimation_training.py", line 182, in weights_output3, biases_output3) File "ex_dnn_head_pose_estimation_training.py", line 105, in multilayer_model _input0_result = tf.matmul(_X, _input0) + _biases_input0 File "/home3/jimmy_extra_space/headpose/deepgaze/.env/local/lib/python2.7/site-packages/tensorflow/python/ops/math_ops.py", line 2018, in matmul a, b, transpose_a=transpose_a, transpose_b=transpose_b, name=name) File "/home3/jimmy_extra_space/headpose/deepgaze/.env/local/lib/python2.7/site-packages/tensorflow/python/ops/gen_math_ops.py", line 4456, in mat_mul name=name) File "/home3/jimmy_extra_space/headpose/deepgaze/.env/local/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 787, in _apply_op_helper op_def=op_def) File "/home3/jimmy_extra_space/headpose/deepgaze/.env/local/lib/python2.7/site-packages/tensorflow/python/util/deprecation.py", line 454, in new_func return func(*args, **kwargs) File "/home3/jimmy_extra_space/headpose/deepgaze/.env/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 3155, in create_op op_def=op_def) File "/home3/jimmy_extra_space/headpose/deepgaze/.env/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1731, in init control_input_ops) File "/home3/jimmy_extra_space/headpose/deepgaze/.env/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1579, in _create_c_op raise ValueError(str(e)) ValueError: Dimensions must be equal, but are 1600 and 4096 for 'MatMul_4' (op: 'MatMul') with input shapes: [2438,1600], [4096,256].

I changed the 1600 to 4096 and was able to proceed with ex_dnn_head_pose_estimation_training.py.

Also I did a brief search here, but didn't find such issue. Hope this is not a duplicate issue.

BR, JimmyYS

NieHw commented 6 years ago

You should resize the face pictures to 64 * 64 at first