Using 2-D input instead of 4-D input matrix #5
Description
Instead of using your feature extractor, we used a different framework for feature extraction. Our h5 file input is in the form of a 2D-matrix(FRAMES, FEATUREDIM) as opposed to the h5 file with 4D-matrix(Frame, Featuredim, Row, Column) generated by feature extraction code of your library. Could you tell us the changes required in your evaluate_ucf101_rgb_ALSTM.py code to process such an input?
THEANO_FLAGS='floatX=float64,device=cpu,mode=FAST_RUN,nvcc.fastmath=True' python evaluate_ucf101_rgb_ALSTM.py is our command entered in terminal.