{"id":20663661,"url":"https://github.com/vita-group/lightgaussian","last_synced_at":"2025-05-15T11:04:36.288Z","repository":{"id":209503865,"uuid":"723803678","full_name":"VITA-Group/LightGaussian","owner":"VITA-Group","description":"[NeurIPS 2024 Spotlight]\"LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPS\", Zhiwen Fan, Kevin Wang, Kairun Wen, Zehao Zhu, Dejia Xu, Zhangyang Wang","archived":false,"fork":false,"pushed_at":"2024-12-30T01:26:10.000Z","size":456,"stargazers_count":675,"open_issues_count":14,"forks_count":65,"subscribers_count":32,"default_branch":"main","last_synced_at":"2025-04-14T19:57:03.185Z","etag":null,"topics":["3d-reconstruction","efficient-inference","gaussian-splatting","neurips-2024","nurips"],"latest_commit_sha":null,"homepage":"https://lightgaussian.github.io/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/VITA-Group.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE.md","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2023-11-26T20:02:23.000Z","updated_at":"2025-04-14T16:17:34.000Z","dependencies_parsed_at":"2024-03-04T00:20:03.609Z","dependency_job_id":"e774fcfd-cb6a-4854-a48a-8c7b9aff1b97","html_url":"https://github.com/VITA-Group/LightGaussian","commit_stats":null,"previous_names":["vita-group/lightgaussian"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/VITA-Group%2FLightGaussian","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/VITA-Group%2FLightGaussian/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/VITA-Group%2FLightGaussian/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/VITA-Group%2FLightGaussian/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/VITA-Group","download_url":"https://codeload.github.com/VITA-Group/LightGaussian/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":254328385,"owners_count":22052632,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["3d-reconstruction","efficient-inference","gaussian-splatting","neurips-2024","nurips"],"created_at":"2024-11-16T19:19:09.006Z","updated_at":"2025-05-15T11:04:36.223Z","avatar_url":"https://github.com/VITA-Group.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPS\n\n\u003cp align=\"center\"\u003e\n\u003ca href=\"https://arxiv.org/abs/2311.17245\"\u003e\u003cimg src=\"https://img.shields.io/badge/Arxiv-2311.17245-B31B1B.svg\"\u003e\u003c/a\u003e\n\u003ca href=\"https://youtu.be/470hul75bSM\"\u003e\u003cimg src=\"https://img.shields.io/badge/Video-Youtube-d61c1c.svg\"\u003e\u003c/a\u003e\n\u003ca href=\"https://lightgaussian.github.io/\"\u003e\u003cimg src=\"https://img.shields.io/badge/Project-Page-048C3D\"\u003e\u003c/a\u003e\n\u003ca href=\"https://github.com/VITA-Group/LightGaussian\"\u003e\u003cimg src=\"https://img.shields.io/github/stars/VITA-Group/LightGaussian\"\u003e\u003c/a\u003e\n\u003c/p\u003e\n\n\u003c!-- [Project Page](https://lightgaussian.github.io) | [Video](https://youtu.be/470hul75bSM) | [Paper](https://lightgaussian.github.io/static/paper/LightGaussian_arxiv.pdf) | [Arxiv](https://arxiv.org/abs/2311.17245) --\u003e\n\u003cdiv\u003e\n\u003cimg src=\"https://lightgaussian.github.io/static/images/teaser.png\" height=\"250\"/\u003e\n\u003c/div\u003e\n\n## User Guidance\n#### Gaussian Prune Ratio, Vector Quantization Ratio vs. FPS, SSIM\n\u003cdiv\u003e\n\u003cimg src=\"https://raw.githubusercontent.com/VITA-Group/LightGaussian/main/static/prune_ratio_vs_ssim.svg\" height=\"250\"/\u003e\n\u003c/div\u003e\n\n#### Mild Compression Ratio, with Minimum Accuracy Degradation\n\u003cdiv\u003e\n\u003cimg src=\"https://github.com/VITA-Group/LightGaussian/blob/main/static/table5.png?raw=true\" height=\"300\"/\u003e\n\u003c/div\u003e\n\n\n## Setup\n#### Local Setup\nThe codebase is based on [gaussian-splatting](https://github.com/graphdeco-inria/gaussian-splatting)\n\nThe used datasets, MipNeRF360 and Tank \u0026 Temple, are hosted by the paper authors [here](https://jonbarron.info/mipnerf360/). \n\nFor installation:\n```\ngit clone --recursive https://github.com/VITA-Group/LightGaussian.git\ncd LightGaussian\n# if you have already cloned LightGaussian:\n# git submodule update --init --recursive\n```\n```shell\nconda env create --file environment.yml\nconda activate lightgaussian\n```\nnote: we modified the \"diff-gaussian-rasterization\" in the submodule to get the Global Significant Score.\n\n\n## Compress to Compact Representation\n\nLightgaussian includes **3 ways** to make the 3D Gaussians be compact\n\u003c!-- #### Option 0 Run all (currently Prune + SH distillation) --\u003e\n\n\n#### Option 1 Prune \u0026 Recovery\nUsers can directly prune a trained 3D-GS checkpoint using the following command (default setting):\n```\nbash scripts/run_prune_finetune.sh\n```\n\nUsers can also train from scratch and jointly prune redundant Gaussians in training using the following command (different setting from the paper):\n```\nbash scripts/run_train_densify_prune.sh\n```\nnote: 3D-GS is trained for 20,000 iterations and then prune it. The resulting ply file is approximately 35% of the size of the original 3D-GS while ensuring a comparable quality level.\n\n\n#### Option 2 SH distillation\nUsers can distill 3D-GS checkpoint using the following command (default setting):\n```\nbash scripts/run_distill_finetune.sh\n```\n\n#### Option 3 VecTree Quantization\nUsers can quantize a pruned and distilled 3D-GS checkpoint using the following command (default setting):\n```\nbash scripts/run_vectree_quantize.sh\n```\n\n\n## Render\nRender with trajectory. By default ellipse, you can change it to spiral or others trajectory by changing to corresponding function.\n```\npython render_video.py --source_path PATH/TO/DATASET --model_path PATH/TO/MODEL --skip_train --skip_test --video\n```\nFor render after the Vectree Quantization stage, you could render them through\n```\npython render_video.py --load_vq\n```\n\n\n## Example\nAn example ckpt for room scene can be downloaded [here](\u003chttps://drive.google.com/drive/folders/1yJeVLQUjYR4cnROOCYuL3o4bXi9atrYH?usp=sharing\u003e), which mainly includes the following several parts:\n\n- point_cloud.ply ——  Pruned, distilled and quantized 3D-GS checkpoint.\n- extreme_saving —— Relevant files obtained after vectree quantization.\n- imp_score.npz —— Global significance used in vectree quantization.\n\n\n\n## TODO List\n- [x] Upload module 1: Prune \u0026 recovery \n- [x] Upload module 2: SH distillation\n- [x] Upload module 3: Vectree Quantization\n- [ ] Upload docker image \n\n## Acknowledgements\nWe would like to express our gratitude to [Yueyu Hu](https://huzi96.github.io/) from NYU for the invaluable discussion on our project.\n\n\n## BibTeX\nIf you find our work useful for your project, please consider citing the following paper.\n\n\n```\n@misc{fan2023lightgaussian, \ntitle={LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPS}, \nauthor={Zhiwen Fan and Kevin Wang and Kairun Wen and Zehao Zhu and Dejia Xu and Zhangyang Wang}, \nyear={2023},\neprint={2311.17245},\narchivePrefix={arXiv},\nprimaryClass={cs.CV} }\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fvita-group%2Flightgaussian","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fvita-group%2Flightgaussian","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fvita-group%2Flightgaussian/lists"}