Practical overview
When it fits
Use Sahi when integrating a supported detection model into a workflow that needs tiled inference on large images, COCO data slicing, or additional inspection of prediction results.
Who it is for
- Python developers building large-image computer-vision inference workflows.
- Computer-vision practitioners evaluating small-object detection results.
- Teams preparing COCO image and annotation data for sliced workflows.
Documented uses
- Apply sliced or standard model prediction to images and video.
- Use supported detection models to detect small objects in large images.
- Slice COCO images and annotations.
- Explore prediction results through the FiftyOne app.
- Inspect outputs with visualization and error-analysis plots.
Requirements & limits
The project is an open-source Python library under the MIT license; its README gives pip as a basic installation route.
- Sahi supplies inference utilities rather than a detection model; select a model and its framework separately.
- The README lists standard and sliced predictions for images and video, plus COCO slicing as a separate utility.
- Framework-specific dependencies differ, and the README gives particular version requirements for some options.
- FiftyOne is used to explore results through a separate app.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install Sahi for a Python project
Install the Sahi library before selecting and installing a supported detection framework.
Before you start
- Python package installer pip
Steps
Run the command below as shown in the repository README.
pip install sahiChoose the detection framework you intend to use and follow its separate installation instructions in the README.
Expected outcome
The README provides the basic Sahi installation path and framework-specific setup options.
Guide limitations
- This covers installation guidance only; it does not establish a working model, inference run or environment compatibility.
- Framework dependencies and version requirements vary by choice.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 10 October 2026
- Original website title
- obss/sahi · Official homepage title · retained in its source language