Video classification using Keras.

Architecture used is CNN+RNN. For CNN, ResNet50 is used. For RNN, GRU is used. Expects a folder with all the video samples and 3 CSV files train.csv, val.csv and test.csv specifying the labels of the training, validation, and test set respectively. The CSVs should have two columns: video_name specifies the video pathname and tag specifies the label of the video. Project includes a custom video data generator to read videos and create batches of samples for the keras' model.fit_generator interface. Latest Cudnn and CudaToolkits are needed. Include data in the correct format and run python train.py.