Rt Percept: Real-Time No-Reference Perceptual Metric Approximation

Rt Percept provides data for the real-time approximation of perceptual metrics without the need for reference or degraded images. Created by Jaliborc and updated in December 2024, the repository supports neural-network-based research into image quality assessment. It is specifically designed for scenarios where a high-quality original image is unavailable for comparison.

Use Cases

• Training neural networks to estimate perceptual metrics without a reference image
• Benchmarking the speed of real-time approximation algorithms for image quality
• Implementing no-reference quality filters in live video pipelines using the provided metrics

Strengths

• Supports no-reference perceptual metric approximation
• Designed for real-time neural network applications
• Updated as of December 2024

Limitations

• Unknown record count and column structure
• Lack of documentation regarding image sources or diversity
• No license information provided in the source repository

Provenance

Source

Jaliborc (GitHub)

Freshness

Updated as of December 20, 2024.