Video-Caption

The second assignment of ADLxMLDS course, NTU 2017 Fall.

Package :

Tensorflow, Numpy, Pandas and other python standard liberary

Techniques:

Seq2seq   s2vt   LSTM   attention  

How to use this project

1. Use extract_feature.py to extract ".avi" to ".npy" on assigned file directory.

   python extract_feature.py [input directory] [output directory] [model] [# of frames]
   

   example: python extract_features.py -i ./Data/peer_review/video -o ./Data/peer_review/feat -m vgg19 -b 80

2. for complete testing model

   python test.py ./Data "s2vt" prediction.txt peer_review
   

Result:

Input a video:

• Video from test dataset folder

Output text:

a man is pouring water from a plastic container.