AMFF-YOLOX: Re-labeled NRSD-MN for Industrial Defect Detection

A re-labeled version of the NRSD-MN dataset specifically for industrial surface defect detection, updated by author chairc in February 2025. It was developed to support the AMFF-YOLOX architecture, which utilizes attention mechanisms and multiple feature fusion for identifying material flaws.

Use Cases

• Training object detection models to identify surface defects using bounding box annotations
• Benchmarking AMFF-YOLOX performance against the original NRSD-MN labels
• Evaluating attention mechanism effectiveness in industrial defect localization

Strengths

• Re-labeled NRSD-MN dataset for improved annotation precision
• Apache-2.0 open-source license
• Directly integrated with AMFF-YOLOX research paper methodology

Limitations

• Narrow domain focus restricted to industrial surface defects
• Unknown record count and class balance
• Lack of detailed column or metadata documentation

Provenance

Source

AMFF-YOLOX: Towards an Attention Mechanism and Multiple Feature Fusion Based on YOLOX for Industrial Defect Detection

Collection Method

Expert re-labeling of the existing NRSD-MN dataset

Freshness

Last updated February 2025.