Image to 3D Datasets

Dataset · Image to 3d

uniocc

Trustworthy Autonomous Systems Laboratory

UniOcc: A Unified Benchmark for Occupancy Forecasting and Prediction in researchers, have you ever been bothered by the fact that popular datasets all have their different formats, and standardizing them is a pain? Have you ever been frustrated by the difficulty of just understanding the file semantics? This challenge is even worse in the occupancy domain. But, UniOcc is here to help. UniOcc is a unified

This dataset contains annotated CAD models with semantic part information for training CAD reverse engineering models. The Img2CAD Annotated CAD Dataset is a comprehensive collection of 3D CAD models with: 4,574 models in total, as reported in the paper; train/test splits follow PartNet. Each cad.h5 file contains: - Part names as keys (e.g., "leg", "top", "back") - Bounding box information ({partname}axisalignedbbox) Img2CAD: Reverse Engineering 3D CAD Models from Images through VLM-Assisted Conditional Factorization This dataset is released under the MIT License. This dataset builds upon the PartNet and ShapeNet datasets. We thank the original authors for their contributions.

Dataset · Image to 3d

uco3d

AI at Meta

This dataset was proposed in UnCommon Objects in 3D.

Alchemy3D-1M is the training corpus for Alchemy3D: source–target 3D editing pairs on PBR-ready assets, with text instructions and image conditions. It contains 1.25M unique assets and 1.38M editing pairs across 7 edit types. Due to Hugging Face storage limits, this repository includes only training metadata, structured latents at resolution 512, and rendered conditions. For GLBs and latents at resolution 1024, please use the ModelScope counterpart Alchemy3D-1M. The dataset contain encoded latents for sparse structure (sslatents), geometry (shapelatents/512), PBR material (pbrlatents/512). csv files are metadata. Latents are packed as partXXX.zip shards. Then unzip all zip files. Apache…

Dataset · Image to 3d

acid-wds

Hong Li

ACID (Aerial Coastline Imagery Dataset) camera trajectories and frames, in the pixelSplat chunk release, repacked as random-access WebDataset shards. Every file in this repository was verified against the source copy (same file set, same byte sizes). WebDataset tar shards (about 3 GB each), one tar row per scene, indexed for random access with ultar.utix files (msgpack rows of entry keys, offsets and sizes, so they can be read without ultar). Every scene row holds: -.tar.utix — the ultar row index (offsets and sizes of every entry). -.tar.centres.bin — float32 camera centres per frame per scene (a small sidecar used for farthest-point view sampling); header DPFV, then a per-scene offset…

Paired RGB images, camera parameters, and complete camera-frustum point clouds. The point clouds contain all available source-mesh vertices inside each image's camera frustum, including vertices hidden behind visible surfaces. They are variable-length XYZ arrays in the OpenCV camera frame (x right, y down, z forward). No fixed point count, sampling, normalization, or patch assignment is applied to the stored targets. - assetsmanifest.json: one record per image, with relative paths, point count, coordinate frame, image derivation modes, and SHA256 hashes. - shards/imgfullgtXXXX.tar: independent archives containing complete sample files, grouped by scene. Each archive is at most 5,000,000,000…

Dataset · Image to 3d

re10k-wds

Hong Li

RealEstate10K camera trajectories and frames, in the pixelSplat chunk release, repacked as random-access WebDataset shards. Every file in this repository was verified against the source copy (same file set, same byte sizes). WebDataset tar shards (about 3 GB each), one tar row per scene, indexed for random access with ultar.utix files (msgpack rows of entry keys, offsets and sizes, so they can be read without ultar). Every scene row holds: -.tar.utix — the ultar row index (offsets and sizes of every entry). -.tar.centres.bin — float32 camera centres per frame per scene (a small sidecar used for farthest-point view sampling); header DPFV, then a per-scene offset table in index row order.…

SAM 3D Objects 多视角微调（reconviagensam3d 分支）所用的训练与评估数据。 每个卷是一个 tar.zst，内部路径 = 原始绝对路径去掉开头的 /。解压到 / 即把 数据放回与原机完全一致的位置，所有符号链接自动生效，无需 relink： 或用仓库里的 restore.sh（支持按分组还原、边解边删控制磁盘占用）。 注意第二行：物体的 images/ 链接走的是 renders/... 这条路径，而 renders 自身 只是一个指向 extracted 的目录级软链接。渲染文件实际都存放在 extracted/ 下， 所以 hypersimrenders 分组里除了那些文件，还显式收录了 renders 这个软链接本身； 少了它，新机上 renders/... 路径不存在，全部 images 链接都会断。 打包用 tar 不加 -h，保留链接本身；链接目标由对应分组整体收录。若展开链接， 同一张 HyperSim 渲染图会被数千个物体各拷一份，实测膨胀约 11 倍。 两条 run 使用相同数据。以 poserelabelAB 为例： 完整参数见各 run 目录下的 launch.sh。 训练用 --pointmap-source gt（默认）时，深度不是从物体目录读，而是由 scripts/hypersimso/rawdepth.py 按 cameras/.json 里的 camname / frameidx 实时拼出原始 HyperSim 路径： HYPERSIMRAWROOT 默认…