Anomaly and Defect Detection: Curated References for Industrial Inspection

This GitHub repository, curated by Howeng98 and last updated in August 2022, serves as a directory of references and descriptions for anomaly, inspection, and defect detection datasets. It categorizes various external data sources specifically focused on manufacturing defects and out-of-distribution detection across multiple industrial domains.

Use Cases

• Benchmarking anomaly detection algorithms by locating datasets with specific 'defect' or 'abnormal' labels
• Sourcing manufacturing-specific image data for 'out-of-distribution-detection' research
• Identifying domain-specific datasets for 'industrial inspection' model training

Strengths

• Categorizes datasets by specific sub-fields like out-of-distribution and defect detection
• Includes 7 distinct tags for filtering manufacturing-related anomalies
• Provides direct references to multiple external inspection sources in one location

Limitations

• Does not host the raw data files directly, requiring users to navigate to external sites
• No standardized schema for the descriptions provided across different entries
• Temporal staleness as it has not been updated since mid-2022

Provenance

Source

GitHub repository by user Howeng98

Collection Method

Manual curation of existing datasets

Freshness

Last updated August 2022; lacks newer datasets released in the subsequent years.