Koldim2001/YOLO-Patch-Based-Inference
Python library providing patch-based inference for YOLO and other Ultralytics deep learning models to improve small object detection and instance segmentation.
Velocity · 7d
+0.7
★ / day
Trend
→steady
star history
This library implements SAHI-like sliced inference for computer vision tasks, enabling detection and segmentation of small objects in images through a patch-based approach. It supports a wide range of Ultralytics models including YOLOv8 through YOLO12, FastSAM, and RTDETR, and provides both standard direct inference and the specialized patch-based workflow. Users can leverage pre-trained models or custom-trained variants with built-in visualization tools.