Egocentric Vision Safetensors

This repository contains safetensors conversions of checkpoints and model assets used by egocentric vision pipelines. It is a mixed-provenance bundle, not a single model release. The conversion to safetensors does not change ownership, copyright, attribution requirements, usage restrictions, or redistribution restrictions from the upstream projects.

Use each file under its own upstream license or terms. When in doubt, download from the original source and review the original license before using the file.

Files and Licenses

File Upstream source Upstream license / terms Notes
droid.safetensors droid.pth from princeton-vl/DROID-SLAM BSD-3-Clause Converted from the official DROID-SLAM checkpoint. Keep the BSD copyright and disclaimer notices when redistributing.
metric3d.safetensors metric_depth_vit_large_800k.pth from JUGGHM/Metric3D, associated with YvanYin/Metric3D BSD-2-Clause; the upstream model card also lists bsd-2-clause Upstream documentation also asks users to contact the authors for commercial inquiries. Review the upstream card and repository before commercial use.
hawor.safetensors hawor/checkpoints/hawor.ckpt from ThunderVVV/HaWoR / GitHub CC-BY-NC-ND-4.0 Non-commercial use only. Do not distribute modified/derived versions. Attribute HaWoR and keep the license notice.
infiller.safetensors hawor/checkpoints/infiller.pt from ThunderVVV/HaWoR / GitHub CC-BY-NC-ND-4.0 Same HaWoR model terms: non-commercial only, no distributed derivatives, attribution required.
yolo.safetensors external/detector.pt as referenced by HaWoR, originally hosted in the WiLoR Space and documented by rolpotamias/WiLoR WiLoR models are documented as CC-BY-NC-ND-4.0; WiLoR also depends on Ultralytics under its own terms Treat this detector as non-commercial/no-derivatives unless you have clearer upstream permission. Also comply with applicable Ultralytics terms for software use.
vggt_omega_1b_512.safetensors vggt_omega_1b_512.pt from facebook/VGGT-Omega CC-BY-NC-4.0 per upstream model card Non-commercial use only, attribution required. The upstream model is gated on Hugging Face; users should comply with upstream access conditions as well as the license.
MANO_RIGHT.safetensors MANO_RIGHT.pkl from a public mirror; official source is the MANO website MANO Software Copyright License for non-commercial scientific research purposes The MANO license restricts use to non-commercial purposes and says the data/software may not be made available to third parties without prior written permission from Max Planck. Use only if you have the required MANO rights.
MANO_LEFT.safetensors MANO_LEFT.pkl from a public mirror; official source is the MANO website MANO Software Copyright License for non-commercial scientific research purposes Same MANO restrictions as MANO_RIGHT.safetensors. The official route is to create a MANO account, accept the license, and download from the MANO website.
manifest.json Manifest generated for this safetensors bundle Repository metadata, not a model license Lists tensor names, hashes, and provenance for droid, metric3d, and MANO conversions.
manifest.hawor-yolo.json Manifest generated for this safetensors bundle Repository metadata, not a model license Lists tensor names, hashes, and provenance for HaWoR and YOLO detector conversions.

Important Restrictions

  • This repository has no single permissive license. The most restrictive relevant upstream terms apply to each file.
  • Non-commercial files must not be used for commercial products, services, paid inference, commercial training, or other commercial purposes unless the upstream licensor grants separate permission.
  • Files under CC-BY-NC-ND-4.0 may be shared for non-commercial purposes with attribution, but modified or derived versions may not be distributed.
  • The MANO files have especially strict redistribution terms. If you do not have written permission to redistribute MANO assets, do not mirror or publish those files; direct users to the official MANO download instead.
  • This model card is a provenance and license summary, not legal advice.

Attribution

If you use any files from this repository, cite and attribute the relevant upstream projects:

  • DROID-SLAM: Zachary Teed and Jia Deng, "DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras."
  • Metric3D / Metric3Dv2: Wei Yin, Mu Hu, and collaborators.
  • HaWoR: Jinglei Zhang, Jiankang Deng, Chao Ma, and Rolandos Alexandros Potamias.
  • WiLoR: Rolandos Alexandros Potamias, Jinglei Zhang, Jiankang Deng, and Stefanos Zafeiriou.
  • VGGT-Omega: Jianyuan Wang, Minghao Chen, Shangzhan Zhang, Nikita Karaev, Johannes Schonberger, Patrick Labatut, Piotr Bojanowski, David Novotny, Andrea Vedaldi, and Christian Rupprecht.
  • MANO: Javier Romero, Dimitrios Tzionas, and Michael J. Black, "Embodied Hands: Modeling and Capturing Hands and Bodies Together."

Original Sources

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