Skip to content

Quickstart

Point Argus at a dataset root. Argus detects YOLO, COCO, and mask layouts automatically.

1. List datasets under a directory

argus-cv list --path /datasets

You will get a table with format, task, classes, and splits.

2. Inspect class balance and background images

argus-cv stats /datasets/traffic

This prints per-class counts per split and a summary line with image totals.

3. Visual inspection

argus-cv view /datasets/traffic --split val

Controls inside the viewer:

  • N or right arrow: next image
  • P or left arrow: previous image
  • Mouse wheel: zoom
  • Drag: pan when zoomed
  • R: reset zoom
  • Q or Esc: quit

4. Split an unsplit dataset

argus-cv split /datasets/traffic -o /datasets/traffic_splits -r 0.8,0.1,0.1

This writes the split dataset to the output path and prints counts for each split.

5. Merge split dataset back to unsplit (optional)

argus-cv unsplit /datasets/traffic_splits -o /datasets/traffic_unsplit

6. Convert formats (optional)

# mask -> YOLO segmentation
argus-cv convert -i /datasets/traffic_masks -o /datasets/traffic_yolo --to yolo-seg

# YOLO segmentation -> COCO
argus-cv convert -i /datasets/traffic_yolo -o /datasets/traffic_coco --to coco