Thinking Hallucination for Video Captioning (THVC)

Official PyTorch implementation for the paper: "Thinking Hallucination for video captioning" accepted at ACCV 2022.

👉 You can find the camera-ready paper here

DOI

Instructions to run the code

1. Install packages mentioned in requirements.txt. Format: pip install package-name

2. Download MSVD.zip, pycocoevalcap.zip, saved_model.pt and put them inside the 'Code and Data' folder and extract. The download links are provided below.

   MSVD.zip - https://drive.google.com/file/d/10Kloc0tOCftEgJ7HgUGnYORiWhmHxnej/view?usp=sharing

   pycocoevalcap.zip - https://drive.google.com/file/d/1BDC5UeqX-0Q0SuqRD_9mWiBSbH5H896H/view?usp=sharing

   saved_model.pt - https://drive.google.com/file/d/1qZaqyWAH6zxq6KjWiHdFpKEffc57vlzj/view?usp=sharing

3. Change the Hyperparameters in config.py (Default hyperparameter is set for the best result)

4. Run Thinking_Hallucination_for_Video_Captioning.ipynb