samouha
commented
6 years ago Probably, this error is caused by the backend used, it is 'Theano' and not TensorFlow ! change it in the file json.keras: (Keras with Theano backend) $ sudo gedit $HOME/.keras/keras.json
Recently,I want to study the project of activitynet-2016-cvprw,but i meet some problem!
As https://github.com/imatge-upc/activitynet-2016-cvprw/issues/18 deccribing, I change the code
def run_all_pipeline(input_video, smoothing_k, activity_threshold): input_size = (112, 112)
to
def run_all_pipeline(input_video, smoothing_k, activity_threshold): input_size = ((16, 112, 112, 3)
, then,i run command "python scripts/run_all_pipeline.py -i changge.mp4",but another probelm appear. as u see blow:
Using TensorFlow backend. Reading Video... Duration: 30.7s FPS: 24.9 Number of frames: 765 Traceback (most recent call last): File "scripts/run_all_pipeline.py", line 284, in
args.activity_threshold)
File "scripts/run_all_pipeline.py", line 35, in run_all_pipeline
video_array = video_array.reshape((nb_clips, length, 3, 112, 112))
ValueError: cannot reshape array of size 4042752 into shape (47,16,3,112,112)
how i can run this demo smoothly? hope for you answer.