BoxLift models

ONNX models used by BoxLift (Chrome extension: draw a 2D box on a camera image, get a 3D cuboid from LiDAR) and by its web demo. Everything runs in the browser through onnxruntime-web.

File	What	Licence
`yolox_s.onnx`	YOLOX-S, COCO-trained 2D detector (official export, Megvii)	Apache 2.0
`heading.onnx`	front/back model, MobileNetV3-small fine-tuned on nuScenes-mini crops	trained on CC BY-NC-SA 4.0 data: non-commercial use
`frustum_seg.onnx`, `frustum_box.onnx`	Frustum PointNets-style 2D box to 3D box, trained on nuScenes and KITTI	trained on CC BY-NC-SA 4.0 / 3.0 data: non-commercial use
`mtrans.onnx`	MTrans (image + LiDAR to 3D box, cars), inference-only export; code MPL-2.0	trained on CC BY-NC-SA data: non-commercial use

SHA256SUMS lists the checksums; frustum.json holds the class sizes used by the frustum models. Details: https://github.com/Gnoltd/boxlift/blob/main/extension/THIRD_PARTY.md

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support