Montgomery
Native object detection, instance segmentation and image classification in Rust with Burn.
View on GitHub →Why Montgomery?
- Native computer vision inference on CPU or GPU
- WGPU training with validation and resumable checkpoints
- Object detection, instance segmentation and image classification
- No Python, no PyTorch, no CUDA, it just works
$ montgomery train --architecture yolo26n --data dataset.yaml --epochs 100
$ montgomery predict --model best.bpk --source image.jpg --json
$ montgomery export-onnx --model best.bpk
Supported Models
| Model Family | Variants | Tasks | Status |
|---|---|---|---|
| YOLOX | nano, tiny, s, m, l, x | Detection | Stable |
| YOLOv3 | tinyu | Detection | Experimental |
| YOLOv8 | n, s, m, l, x | Detection, segmentation, classification | Experimental |
| YOLOv10 | n, s, m, b, l, x | Detection | Experimental |
| YOLO11 | n, s, m, l, x | Detection, segmentation, classification | Experimental |
| YOLO12 | n, s, m, l, x | Detection | Experimental |
| YOLO26 | n, s, m, l, x | Detection, segmentation, classification | Experimental |