ABSTRACT:

This project addresses the challenge of video captioning by implementing an encoder-decoder architecture with Long Short-Term Memory (LSTM) networks to automatically generate text descriptions of video content. Using the Microsoft Video Description (MSVD) dataset, video frames are processed through a pre-trained VGG16 model to extract features, which are then passed to the LSTM-based model. The system is trained to predict captions with a limited vocabulary of 1,500 words. By employing greedy search for inference, the model can generate coherent captions, offering applications in video search, recommendation systems, and content accessibility.


OBJECTIVE:

The objective of this project is to develop a deep learning-based video captioning system that automatically generates textual descriptions for video content. Specifically, the system aims to:
1. Extract meaningful features from video frames using a pre-trained VGG16 model.
2. Implement an encoder-decoder architecture with Long Short-Term Memory (LSTM) networks to process sequential video frames and generate captions.
3. Train the model on the Microsoft Video Description (MSVD) dataset and evaluate its ability to produce coherent and contextually relevant captions for unseen videos.
4. Explore different search strategies (e.g., greedy search) during inference to improve the quality of generated captions.
5. Demonstrate the potential applications of the model in video search, recommendation systems, and improving video accessibility.