Skip to content

Argus

Vision AI dataset work, without the friction.

Argus is a focused CLI for YOLO, COCO, and semantic mask datasets. List datasets, inspect class balance, view annotations, and move between split and unsplit layouts cleanly for training.

Get started See the commands 
# Discover datasets
argus-cv list --path /data

# Instant class stats
argus-cv stats /data/animals

# Visual inspection
argus-cv view /data/animals --split val

# Convert mask dataset to YOLO segmentation
argus-cv convert -i /data/animals_masks -o /data/animals_yolo --to yolo-seg

Format-aware

Detects YOLO, COCO, and mask datasets by structure and metadata.

Readable statistics

Per-class counts and background image totals at a glance.

Annotation viewer

Browse images with boxes, polygons, and mask overlays. Pan and zoom included.

Clean splits

Stratified splitting for YOLO and COCO with deterministic seeds.

Common workflows

argus-cv list --path /datasets
argus-cv stats /datasets/retail

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

argus-cv split /datasets/retail -o /datasets/retail_splits -r 0.8,0.1,0.1

argus-cv unsplit /datasets/retail_splits -o /datasets/retail_unsplit

argus-cv convert -i /datasets/retail_masks -o /datasets/retail_yolo --to yolo-seg

What Argus expects

Argus detects datasets by their structure. If you point it at the dataset root, commands will usually just work.

  • YOLO: a data.yaml (or .yml) with names, plus images/ and labels/
  • COCO: one or more instances_*.json files in annotations/

Head to the formats guide for the full layouts and edge cases.

Explore dataset formats

Library usage

Using Argus as a Python library? See the API reference for dataset classes and core operations.

Open Python API reference