#6060th of 68 in Training and fine-tuning
ml-hypersim
Synthetic indoor-scene dataset with per-pixel labels, plus a V-Ray rendering toolkit
- Stars
- 2.1k
- License
- custom license
- Last commit
- Sep 2026
- Language
- Python
1,993 on 25 Apr 20262,075 on 14 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Hypersim is a photorealistic synthetic dataset of 77,400 images across 461 indoor scenes (74,619 in the public release), with depth, normals, semantic instance labels, 3D bounding boxes and camera poses. Images are split into diffuse reflectance, diffuse illumination and a non-diffuse residual. The repo also includes the Hypersim Toolkit, Python tools for generating similar datasets from V-Ray scenes.
Who it is for: Researchers training scene-understanding models on synthetic indoor data
Strengths
- Per-pixel depth, normals, NYU40 semantics, instances and 9-DOF boxes for every image
- Lossless HDR HDF5 images with color = reflectance * illumination + residual
- Standard scene-level train/val/test split (v1) included as CSV
- Toolkit can generate new labeled datasets from V-Ray scenes
Weaknesses
- Full dataset is roughly 1.9TB across a few hundred ZIP files
- Ground truth meshes require purchasing scene assets from TurboSquid
- Toolkit depends on V-Ray, which is under Chaos Group's own terms
- Some asset reuse across scenes leaks objects between train and test splits
What it needs
- Needs V-Ray Standalone, V-Ray AppSDK, curl, unzip, Python
Also in Training and fine-tuning
See all 68| Rank | Project | Score |
|---|---|---|
| 1 | peftPython library for LoRA and other parameter-efficient fine-tuning of Transformers models | 89 out of 100 |
| 2 | unslothDesktop and web app to run and fine-tune local models | 88 out of 100 |
| 3 | trlPost-training library with SFT, DPO, GRPO and KTO trainers | 88 out of 100 |
| #4 | deepspeedPyTorch library for distributed training of very large models with ZeRO | 87 out of 100 |
| #5 | megatron-lmDistributed training library for large transformer models on NVIDIA GPUs | 86 out of 100 |