---

Video GPT Project

Overview

Welcome to the Video GPT repository. This project focuses on video caption generation using advanced models and techniques to process videos efficiently. Our system is built to handle large video files by reducing and eliminating redundant frames, leveraging models like InternV2, and generating summarized text from the remaining frames.

Demo Link

You can access the prototype of Video GPT project here : Demo Link

Prototype

Project Structure

The repository is organized into the following main folders:

• development/: Contains the core functionalities under active development.

  • frame_reduction/: Modules to reduce video frames.
  • redundant_frame_elimination/: Scripts to eliminate redundant frames from videos.
  • internv2_model/: Implementation of the InternV2 model for processing redundant frames.
  • text_summarization/: Summarizes text based on the processed frames.

• production/: Holds the web interface and all code ready for deployment.

  • web_interface/: The front-end and back-end for the Video GPT web application.

• research/: Contains experimental scripts and files for ongoing research and model improvements.

  • models_experiments/: Files related to exploring and refining video-to-text models.

Key Features

• Frame Reduction: Efficiently reduces the number of frames in a video to optimize processing time.
• Redundant Frame Elimination: Identifies and removes frames that do not add value to the video's context.
• InternV2 Model: Utilizes the InternV2 model for redundant frame analysis and processing.
• Text Summarization: Automatically generates text summaries from the optimized video frames.

Frame Reduction Idea Summary

video-gpt/
│
├── development/
│   ├── frame_reduction/                  # Reducing video frames functionality
│   ├── redundant_frame_elimination/      # Eliminating redundant frames functionality
│   ├── internv2_model/                   # InternV2 model for redundant frames
│   ├── text_summarization/               # Text summarization for redundant frames
│   
├── production/
│   ├── web_interface/                    # Web interface for Video GPT project
│   
├── research/
│   ├── models_experiments/               # Research files and model experiments
│  
├── requirements.txt                      # Python dependencies
|
├── index.html                            # Web Interface
│
├── README.md                             # Project README file
│
└── .gitignore                            # Git ignore file

This structure integrates both the Video GPT features and the organization of components like development, production, and research folders.

Setup Instructions

1. Clone the repository:

   git clone https://github.com/yourusername/videogpt.git

2. Install the necessary dependencies:

   cd videogpt
   pip install -r requirements.txt

3. Follow the respective folder README files for setting up individual components.

Usage

Development

For development purposes, you can start by exploring the individual functionalities under the development/ folder. Each module comes with its own README and usage instructions.

Production

To run the web interface, navigate to the production/web_interface/ folder, and follow the setup guide to launch the Video GPT web application.

Research

All ongoing experiments and research efforts are located under the research/ folder. This includes experiments with different models and optimization techniques.

Contributing

Feel free to submit pull requests or open issues for any bugs, feature requests, or improvements. We welcome contributions from the community!

---