Formula used
width = xmax − xmin
height = ymax − ymin
area = width × height
relative area = box area ÷ image area
aspect ratio = width ÷ height
coverage percentage = relative area × 100
height = ymax − ymin
area = width × height
relative area = box area ÷ image area
aspect ratio = width ÷ height
coverage percentage = relative area × 100
How to use
- Paste annotations or upload a supported data file.
- Select the coordinate format and matching column positions.
- Choose validation, filters, outliers, and anchor settings.
- Submit the form to calculate every box metric.
- Explore charts, tables, insights, anchors, and export options.
Example data structure
| image_id | class | image_width | image_height | a | b | c | d | confidence | split |
|---|---|---|---|---|---|---|---|---|---|
| img_001.jpg | person | 1280 | 720 | 110 | 80 | 310 | 650 | 0.98 | train |
Frequently asked questions
Which coordinate formats are supported?
Use corner, top-left size, center size, or YOLO coordinates.
Can normalized coordinates be analyzed?
Yes. Image dimensions convert normalized values into pixel measurements.
How are tiny objects identified?
Custom thresholds classify boxes using relative or pixel area.
What makes an annotation invalid?
Missing values, impossible dimensions, reversed corners, and boundary violations.
How are anchor boxes recommended?
K-means groups box widths and heights using selected distance rules.
What does average anchor IoU mean?
It measures how closely recommended anchors match observed box shapes.
Can I compare training and validation data?
Yes. Use split charts and filter fields for comparisons.
Can outliers be removed?
Yes. Detect them first, then exclude them from analysis.
Which exports are available?
Export charts, processed rows, summaries, outliers, anchors, and reports.