search

eldar / pose-tensorflow

Human Pose estimation with TensorFlow framework
GNU Lesser General Public License v3.0
1.14k stars 384 forks source link

TypeError: resnet_v1_101() got an unexpected keyword argument 'is_training' #53

Open galaxy-fangfang opened 6 years ago

galaxy-fangfang commented 6 years ago

i was testing the demo for single person detection when i encountered this error. i installed tensorflow for python3 for virtualenv environment. i don't know if the error was caused by my tensorflow or by the code. please help me with this error, thanks!

(tensorflow)fangfang@dl:~/PycharmProjects/pose-tensorflow$ TF_CUDNN_USE_AUTOTUNE=0 python3 demo/singleperson.py I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcublas.so.8.0 locally I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcudnn.so.5 locally I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcufft.so.8.0 locally I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcuda.so.1 locally I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcurand.so.8.0 locally Traceback (most recent call last): File "demo/singleperson.py", line 17, in sess, inputs, outputs = predict.setup_pose_prediction(cfg) File "/home/fangfang/PycharmProjects/pose-tensorflow/nnet/predict.py", line 11, in setup_pose_prediction outputs = pose_net(cfg).test(inputs) File "/home/fangfang/PycharmProjects/pose-tensorflow/nnet/pose_net.py", line 89, in test heads = self.get_net(inputs) File "/home/fangfang/PycharmProjects/pose-tensorflow/nnet/pose_net.py", line 85, in get_net net, end_points = self.extract_features(inputs) File "/home/fangfang/PycharmProjects/pose-tensorflow/nnet/pose_net.py", line 55, in extract_features net, end_points = net_fun(im_centered, global_pool=False, output_stride=16, is_training=False) TypeError: resnet_v1_101() got an unexpected keyword argument 'is_training'

eldar commented 6 years ago

Are you using Tensorflow 1.4? If not, please upgrade.

galaxy-fangfang commented 6 years ago

@eldar thank you for your reply. by the way, i also want to know if i train on my own dataset, what is the sequence of the joints in your example? or it is just defined by myself? if i want to train mutil_person, does the dataset mat need to be somewhat different?

joints = {[ ... 0, 175, 261; ... % 0-indexed joint ID, X coordinate, Y coordinate 1, 173, 178; ... 2, 144, 122; ... 3, 193, 124; ... ]};