The code for the paper “Domain-Specific Semantics Guided Approach to Video Captioning”, WACV 2020.

• Python 3.5
• Keras built with tensorflow backend [Version 2.31]
• Tensflow [Version 1.15.0]

• Resnet-152 and C3D Features are provided in the data folder
• For semantic features we have provided Ground truth from SCN paper. You can train your own MLP to extract features
• You can evaluate the model using MSCOCO pycocotools

Please cite our WACV paper if it helps your research:

@INPROCEEDINGS{Domain_vid_capt_wacv_2020,
author={M. {Hemalatha} and C. C. {Sekhar}},
booktitle={2020 IEEE Winter Conference on Applications of Computer Vision (WACV)},
title={Domain-Specific Semantics Guided Approach to Video Captioning},
year={2020},
pages={1576-1585}, }