tiange/Cap3D
This repository hosts data for Scalable 3D Captioning with Pretrained Models and View Selection for 3D Captioning via Diffusion Ranking, including descriptive captions for 3D objects in Objaverse, Objaverse-XL, ABO, and ShapeNet. This repo also includes point clouds and rendered images with camera, depth, and MatAlpha information of Objaverse objects, as well as their Shap-E latent codes. All the captions and data provided by our papers are released under ODC-By 1.0 license. Important… See the full description on the dataset page: https://huggingface.co/datasets/tiange/Cap3D.
Thành tựu nổi bật
Những danh hiệu mới nhất của dự án
#1 Repository Của Tuần
#1 Repository Của Tuần
#1 Repository Của Tuần
🧊 Cap3D: Bộ dữ liệu chú thích 3D quy mô lớn cho mô hình pretrained
Hơn 1,6 triệu cặp 3D – caption, kèm point cloud, ảnh render và mã latent.
Cap3D là bộ dữ liệu phục vụ hai nghiên cứu của nhóm nghiên cứu 3D vision: Scalable 3D Captioning with Pretrained Models và View Selection for 3D Captioning via Diffusion Ranking. Dữ liệu bao phủ bốn nguồn lớn — Objaverse, Objaverse-XL, ABO và ShapeNet — bao gồm chú thích mô tả chi tiết từng mô hình 3D, cùng point cloud, ảnh render (với thông tin camera, độ sâu, mặt nạ MatAlpha) và mã latent của Shap-E.
✨ Điểm nổi bật
- 1,604,291 cặp 3D – caption: Sinh tự động bởi Cap3D và DiffuRank cho Objaverse và Objaverse-XL.
- Point cloud màu sắc: 16,384 điểm RGB gắn liền với từng mô hình, lưu dưới định dạng
.pt(tensor[16384, 6]) và.ply. - Ảnh render chi tiết: 20 góc ảnh mỗi mô hình, kèm camera (intrinsic/extrinsic), depth và mask.
- Mã latent Shap-E: Hỗ trợ trực tiếp cho các tác vụ sinh 3D từ văn bản hoặc hình ảnh.
- Hỗ trợ nhiều nguồn dữ liệu: Bao gồm cả ABO và ShapeNet, đáp ứng nhu cầu đa domen.
- Giấy phép ODC-By 1.0: Dữ liệu được công bố công khai, yêu cầu ghi công nguồn.
💡 Vì sao đáng chú ý
Cap3D khác biệt nhờ quy mô khổng lồ và sự đa dạng của dữ liệu — không chỉ dừng ở caption mà còn cung cấp đầy đủ các dạng biểu diễn trực quan như point cloud và ảnh render. Điều này giúp các nhà nghiên cứu huấn luyện mô hình 3D hiệu quả hơn, đặc biệt trong các bài toán text-to-3D và image-to-3D.
#3DCaptioning #TextTo3D #ImageTo3D #Objaverse #ShapeNet #PointCloud #TiniX
Dự án tương tự
Khám phá các dự án và mô hình AI khác có sự tương đồng.
Objaverse-XL "> Objaverse-XL is an open dataset of over 10 million 3D objects! With it, we train Zero123-XL, a foundation model for 3D, observing incredible 3D generalization abilities: 🧵👇 Scale Comparison Objaverse 1.0 was released back in December. It was a step in the right direction, but still relatively small with 800K objects. Objaverse-XL is over an order of magnitude larger and much more diverse! Unlocking Generalization… See the full description on the dataset page: https://huggingface.co/datasets/allenai/objaverse-xl.
Objaverse Objaverse is a Massive Dataset with 800K+ Annotated 3D Objects. More documentation is coming soon. In the meantime, please see our paper and website for additional details. License The use of the dataset as a whole is licensed under the ODC-By v1.0 license. Individual objects in Objaverse are all licensed as creative commons distributable objects, and may be under the following licenses: CC-BY 4.0 - 721K objects CC-BY-NC 4.0 - 25K objects CC-BY-NC-SA… See the full description on the dataset page: https://huggingface.co/datasets/allenai/objaverse.
Zero-to-CAD 1M One million executable, interpretable CAD construction sequences synthesized entirely without real-world data. Zero-to-CAD: Agentic Synthesis of Interpretable CAD Programs at Million-Scale Without Real Data Mohammadmehdi Ataei, Farzaneh Askari, Kamal Rahimi Malekshan, Pradeep Kumar Jayaraman Autodesk Research Related Resources Resource Link 📄 Paper Zero-to-CAD: Agentic Synthesis of Interpretable CAD Programs at… See the full description on the dataset page: https://huggingface.co/datasets/ADSKAILab/Zero-To-CAD-1m.
HY3D-Bench Project page | Paper | Code 🔥 News Feb 04, 2026: 🎉 We release HY3D-Bench - a comprehensive collection of high-quality 3D datasets with 3 DIFFERENT subsets! ✨ Full-level Dataset: 252K+ watertight meshes with multi-view renderings and sampled points ✨ Part-level Dataset: 240K+ objects with fine-grained part decomposition ✨ Synthetic Dataset: 125K+ AI-synthesized objects across 1,252 categories 🚀 Baseline Model:… See the full description on the dataset page: https://huggingface.co/datasets/tencent/HY3D-Bench.
3DCodeVerse — Towards Building the Universe via Code 3DCodeVerse aims to gather any code that creates, renders, or governs 3D — every dialect, one corpus. Text / Image / Video ↔ 3D Code ↔ 3D Assets Every entry is a self-contained, runnable project: the code, its rendered outputs, a metadata card, and text captions. Domains in scope: Object Modeling · Scene & World Modeling · Rendering · Materials & Shaders · Physics & Simulation · Animation & Articulation. 📄 Paper:… See the full description on the dataset page: https://huggingface.co/datasets/ilabai/3dcodeverse.
TRELLIS-500K TRELLIS-500K is a dataset of 500K 3D assets curated from Objaverse(XL), ABO, 3D-FUTURE, HSSD, and Toys4k, filtered based on aesthetic scores. This dataset serves for 3D generation tasks. It was introduced in the paper Structured 3D Latents for Scalable and Versatile 3D Generation. Dataset Statistics The following table summarizes the dataset's filtering and composition: NOTE: Some of the 3D assets lack text captions. Please filter out such assets if… See the full description on the dataset page: https://huggingface.co/datasets/JeffreyXiang/TRELLIS-500K.
SpatialVID: A Large-Scale Video Dataset with Spatial Annotations Jiahao Wang1* Yufeng Yuan1* Rujie Zheng1* Youtian Lin1 Jian Gao1 Lin-Zhuo Chen1 Yajie Bao1 Yi Zhang1 Chang Zeng1 Yanxi Zhou1 Xiaoxiao Long1 Hao Zhu1 Zhaoxiang Zhang2 Xun Cao1 Yao Yao1† 1Nanjing University 2Institute of Automation, Chinese Academy of Science *Equal Contribution †Corresponding Author CVPR 2026… See the full description on the dataset page: https://huggingface.co/datasets/SpatialVID/SpatialVID.
Project page Paper Code 3dcodebench.com arXiv:2606.01057 gaoypeng/3dcodebench News [06/01/2026] Paper released on arXiv: 3DCodeBench: Benchmarking Agentic Procedural 3D Modeling Via Code. Note. This is an open-source reproduction of 3DCodeBench. ⚠️ Under final check. The 3DCodeData/ code is still undergoing final quality review and may contain occasional issues (non-executable scripts, mismatched captions/renders, or imperfect geometry). If you run… See the full description on the dataset page: https://huggingface.co/datasets/YipengGao/3DCode.
SpatialVID: A Large-Scale Video Dataset with Spatial Annotations Jiahao Wang1* Yufeng Yuan1* Rujie Zheng1* Youtian Lin1 Jian Gao1 Lin-Zhuo Chen1 Yajie Bao1 Yi Zhang1 Chang Zeng1 Yanxi Zhou1 Xiaoxiao Long1 Hao Zhu1 Zhaoxiang Zhang2 Xun Cao1 Yao Yao1† 1Nanjing University 2Institute of Automation, Chinese Academy of Science *Equal Contribution †Corresponding Author CVPR 2026… See the full description on the dataset page: https://huggingface.co/datasets/FelixYuan/SpatialVID-HQ.