Super Resolution for Low-Quality Videos

This project applies super-resolution to low-quality videos, enhancing their resolution using deep learning models. It uses the Real-ESRGAN architecture to upscale videos to a higher resolution (e.g., 1440p), improving their visual quality.

Features

• Super Resolution: Enhances video frames from lower resolutions to 1440p using the Real-ESRGAN model.
• Deep Learning Model: Utilizes a pre-trained Real-ESRGAN model built on the RRDBNet architecture.
• GPU Support: Optimized to run on GPUs (CUDA) if available, with a fallback to CPU.
• Progress Tracking: Displays a progress bar for each frame processed, providing real-time updates on the video enhancement process.
• Input/Output Support: Processes input videos in formats like MP4 and outputs the enhanced video in higher resolution.

How It Works

1. Model Architecture: The model uses the RRDBNet architecture, which is loaded with pre-trained weights from the Real-ESRGAN repository.
2. Frame Processing: Each video frame is converted to RGB, passed through the upscaling model, and then written to the output video at the desired resolution (1440p).
3. Video Enhancement: The video is processed frame by frame, applying super-resolution, and then saving the results in a higher-quality output file.

Technologies Used

• Deep Learning Model: Real-ESRGAN based on RRDBNet.
• Computer Vision: OpenCV for reading and writing video frames.
• GPU Acceleration: Supports CUDA for faster video processing.
• Python Libraries: Torch, TQDM, NumPy, OpenCV.

Setup and Installation

1. Clone the repository and install the required dependencies:

   pip install -r requirements.txt