Skip to main content

Dataset Health Check (Image Detection)

Image Detection​

Overview Tab​

High-level summary: total images, annotated count, issues found, duplicates or corrupted files detected.

Dataset Health Check overview

Quality Checks Tab​

CheckWhat it detects
Small boxesBounding boxes smaller than 0.5% of image area
Low resolutionImages below 640×480 px
Box count anomaliesUnusually high or low bounding box count per image
Missing class labelsClass IDs not present in classes.txt
Class imbalanceClasses with significantly more/fewer examples

Actions Tab​

  • Duplicate image management — side-by-side previews of detected duplicates
  • Unannotated image management — lists images with no label file
  • Soft-delete — marks images as excluded from training without deleting from disk