Industrial Defect Segmentation Datasets

14 industrial inspection datasets are curated from Roboflow for vision-based manufacturing analysis. The collection includes polygon annotations derived from bounding box labels to enable precise defect segmentation. The datasets were compiled by VISION-Workshop and last updated in June 2023.

Use Cases

• Train semantic segmentation models using polygon annotations to identify defect regions on manufactured parts.
• Benchmark object detection algorithms on bounding box labels across diverse industrial environments.
• Develop anomaly detection systems for specific manufacturing processes using the curated collection of 14 datasets.
• Evaluate model generalization across different materials and industries covered by the dataset collection.

Strengths

• Collection of 14 distinct industrial inspection datasets.
• Datasets include polygon annotations for precise segmentation.

Limitations

• Specific row counts, image resolutions, and class distributions for individual datasets are not provided.
• The datasets are curated from a single source (Roboflow), which may limit diversity.

Provenance

Source

Curated from Roboflow.

Collection Method

Datasets were collected and annotated with polygon labels based on provided bounding boxes.

Freshness

Last updated in June 2023.