AutoVision

Enterprise-grade visual model training framework: self-supervised pretrain / distill → finetune → evaluate → export in one chain. Covers detection, rotated boxes, instance / panoptic / semantic segmentation, pose, depth, and classification; the Python API and autovision-cli share the same strict Pydantic validation.

Distribution name: the PyPI package is autovision-framework, while the import package stays autovision and the CLI stays autovision-cli. This project is unrelated to the unrelated autovision 0.0.1 package on PyPI; make sure you install autovision-framework.

Links: Source repository · Documentation · Changelog

Installation

Python ≥ 3.10:

pip install autovision-framework

# Common extras: HF backbones + timm + ONNX export
pip install "autovision-framework[transformers,timm,export]"

# Pretrained RF-DETR direct inference
pip install "autovision-framework[rfdetr,export]"

Model packages load lazily per extra; a missing extra yields an empty registry listing instead of an error. See the repository README for the full extras table.

Quick start

No training — load released weights and predict:

import autovision

model = autovision.detection("rfdetr/rf-detr-small", pretrained=True)
result = model.predict("image.jpg")   # {'bboxes': ..., 'scores': ..., 'labels': ...}
model.export("model.onnx")            # parity gate + postprocess spec

Object detection, train → validate → predict → export:

import autovision

data = autovision.ensure_coco8()   # synthetic smoke dataset; swap in COCO/YOLO configs
model = autovision.detection("rtdetr/resnet18").train(
    out="runs/det1", data=data, steps=2000, batch_size="auto",
)
model.val(data=data)          # {'map': ..., 'map_50': ...}
model.predict("image.jpg")    # {'bboxes': ..., 'scores': ..., 'labels': ...}
model.export("model.onnx")    # dispatches by .onnx / .engine / .pt suffix

Equivalent CLI (same Pydantic validation):

autovision-cli train-object-detection out=runs/det1 data=data.yaml steps=2000
autovision-cli val     model=runs/det1/exported_models/exported_best.pt data=coco8
autovision-cli predict model=runs/det1/exported_models/exported_best.pt images=dir/ out=preds/ threshold=0.5

Common extras

License

Apache-2.0 (AutoVision original code; full text: https://www.apache.org/licenses/LICENSE-2.0). Third-party component and weight provenance ship in NOTICE and THIRD_PARTY_LICENSES.md. Restricted-license weights (DINOv2 CC-BY-NC 4.0 / DINOv3 License) are not downloaded by default; user-held weights may not be redistributed.

0.3.0 licensing status: the DEIMv2-derived SpatialPriorModule was removed; DINOv3 ViT variants are kept with the no-STA adapter, and the release no longer contains DEIMv2 code. Legacy backbone.sta.* ViT checkpoints are unsupported and must be retrained/re-exported. See docs/provenance/LEDGER.md for details.

Full documentation (model zoo, benchmarks, export details, known limits) is available on the documentation site and in the docs/ directory of the source repository.