obss/sahi

Framework agnostic sliced/tiled inference + interactive ui + error analysis plots

81
/ 100
Verified

Implements adaptive slicing strategies to partition large images into smaller regions for inference, then merges detections with duplicate elimination—enabling efficient small object detection without retraining models. Supports multiple detection frameworks (YOLOv5/v8, MMDetection, HuggingFace Transformers, TorchVision) through a unified API, plus dataset utilities for COCO format slicing and error analysis via FiftyOne integration.

5,160 stars. Used by 1 other package. Actively maintained with 28 commits in the last 30 days. Available on PyPI.

Maintenance 23 / 25
Adoption 11 / 25
Maturity 25 / 25
Community 22 / 25

How are scores calculated?

Stars

5,160

Forks

735

Language

Python

License

MIT

Last pushed

Mar 12, 2026

Commits (30d)

28

Dependencies

11

Reverse dependents

1

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/obss/sahi"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.