360-Degree Video Super-Resolution

A powerful AI-driven solution for enhancing 360-degree videos using Real-ESRGAN and GFPGAN technologies. This project combines advanced super-resolution techniques with intelligent face enhancement to dramatically improve the quality of 360-degree video content.

Features

• 4x Super-Resolution: Upscale 360-degree videos to 4 times their original resolution
• AI-Powered Face Enhancement: Intelligent face detection and enhancement using GFPGAN
• Smart Blending: Advanced blending algorithms for seamless results
• Batch Processing: Process multiple frames efficiently
• Audio Preservation: Maintains original audio quality during enhancement
• Google Colab Ready: Pre-configured Jupyter notebook for easy cloud execution

Quick Start

Option 1: Google Colab (Recommended)

1. Open the Real_ESRGAN_Video.ipynb notebook in Google Colab
2. Run all cells in sequence
3. Upload your 360-degree video when prompted
4. Download the enhanced result

Option 2: Local Installation

# Clone the repository
git clone https://github.com/WWIIITT/360-degree-video-super-resolution.git
cd 360-degree-video-super-resolution

# Install dependencies
pip install -r requirements.txt

# Run the enhancement script
python real_esrgan_video.py

Requirements

Python: 3.9 or higher, GPU: CUDA-compatible GPU recommended for faster processing, Dependencies:

• torch==1.13.1
• torchvision==0.13.1
• numpy==1.23.5
• basicsr==1.4.2
• facexlib==0.3.0
• gfpgan==1.3.8

Installation

Automatic Setup (Google Colab)

The Jupyter notebook automatically handles all installations and dependencies.

Manual Setup

# Install PyTorch (adjust CUDA version as needed)
pip install torch==1.13.1 torchvision==0.13.1

# Install required packages
pip install numpy==1.23.5
pip install basicsr==1.4.2 facexlib==0.3.0 gfpgan==1.3.8

# Install OpenCV
pip install opencv-python opencv-contrib-python

# Clone Real-ESRGAN
git clone https://github.com/xinntao/Real-ESRGAN.git
cd Real-ESRGAN
python setup.py install

# Clone GFPGAN
git clone https://github.com/TencentARC/GFPGAN.git
cd GFPGAN
python setup.py install

Usage

Basic Usage

# Import required modules
from real_esrgan_video import enhance_360_video

# Enhance your 360-degree video
enhanced_video = enhance_360_video(
    input_path="your_360_video.mp4",
    output_path="enhanced_360_video.mp4",
    scale_factor=4
)

Advanced Configuration

# Customize enhancement parameters
enhanced_video = enhance_360_video(
    input_path="input.mp4",
    output_path="output.mp4",
    scale_factor=4,
    face_enhance=True,
    gfpgan_weight=0.2,
    blur_kernel_size=151,
    blending_alpha=0.3
)

Technical Details

Processing Pipeline

1. Frame Extraction: Extract individual frames from input video
2. Super-Resolution: Apply Real-ESRGAN for 4x upscaling
3. Face Detection: Identify faces using Haar cascades
4. Face Enhancement: Apply GFPGAN to detected face regions
5. Smart Blending: Seamlessly blend enhanced faces with upscaled background
6. Post-Processing: Apply CLAHE and noise reduction
7. Video Reconstruction: Reassemble frames with original audio

Supported Formats

• Input: MP4, AVI, MOV, MKV, INSV (360-degree formats)
• Output: MP4 with H.264 encoding
• Audio: AAC encoding at 192kbps

Acknowledgments

• Real-ESRGAN by Xintao Wang et al.
• GFPGAN by Tencent ARC Lab
• BasicSR for the super-resolution framework