OriginLab Game-Depth
Collection
Dense depth from game engines: dataset + Lotus-recipe pretrained and NYU-fine-tuned models. • 2 items • Updated • 9
Request access to Origin Lab Lotus Game-Depth
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
Two access tracks, defined in LICENSE.md. Track A, Internal Evaluation: 90 days to train and evaluate models internally to assess the data. No obligation to publish anything. Track B, Non-Commercial Research: research use with attribution; publish freely. Under both tracks there is no commercial use, no deployment, and no redistribution in any form. Commercial licensing requires a direct agreement with Origin Lab: originlab.ai/hf.
Log in or Sign Up to review the conditions and access this model content.
Origin Lab Lotus Game-Depth (pretrained + NYU fine-tuned)
Website: originlab.ai
Two Lotus-recipe latent-diffusion depth checkpoints (SD2-base UNet, 8-channel conv_in, single-step x0 at t=999, trunc_disparity), in one repo:
pretrained/- trained from scratch on the OriginLab Game-Depth dataset (game-engine z-buffers), no real data. Zero-shot KITTI AbsRel 0.191 (Lotus 0.224, Marigold 0.244).nyu-ft/- the above fine-tuned on real NYU Depth V2. NYU AbsRel 0.116, on par with a fairly-tuned Lotus baseline (0.115) using 0% indoor pretraining data and ~4x fewer frames.
Dataset: https://huggingface.co/datasets/originlab/game-depth
Load
from diffusers import UNet2DConditionModel
# game-pretrained
unet = UNet2DConditionModel.from_pretrained("originlab/lotus-game-depth", subfolder="pretrained/unet")
# NYU fine-tuned
unet = UNet2DConditionModel.from_pretrained("originlab/lotus-game-depth", subfolder="nyu-ft/unet")
Each subfolder contains the full pipeline (unet, vae, text_encoder, scheduler, ...); run with the Lotus single-step depth pipeline.
License
Origin Lab Data License (LICENSE.md). Two tracks; select one
when requesting access.
- Track A, Internal Evaluation. 90 days to train and evaluate models internally in order to assess the data. Origin Lab does not require you to publish or open-source anything you train. Delete at the end, or convert to a commercial agreement.
- Track B, Non-Commercial Research. Research use with attribution. Papers, open weights, and benchmarks are welcome.
Under both tracks: no commercial use, no deployment, and no redistribution of the data or these checkpoints in any form. Any commercial use of the data, or of a model trained on it, requires a direct license from Origin Lab: originlab.ai/hf.
- Downloads last month
- -
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support