The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
This is image dataset for 100 objects. For every object, 514 images are given from viewpoints distributed evenly across a hemisphere. Images are supplemented with laser-scanned meshes (ground truth). Image resolution is 4112 × 3008 pixels.
A sample of 514 instances from a larger dataset is presented. The full dataset contains positions from 12 cameras mounted on an arc, which is rotated in 6 degree steps, resulting in 720 images in total. Because the instances near the pole become progressively closer to each other, 514 instances were selected to maintain approximate equidistance and reduce unnecessary redundancy caused by overlap. This sample is fully representative of the larger set. Each image is provided in JPG and EXR formats. The EXR images offer high dynamic range without tone mapping.
Objects are placed on a rotating disk marked with a 16h5 AprilTag pattern along its perimeter. Camera settings and metadata are embedded into EXIF data for every image.
The selection of objects is designed to represent a diverse range of shapes and materials. For training radiance field models, no specific data splits are recommended - subsets may be chosen randomly.
The dataset is self-contained and free from known errors, noise sources, or redundancies. It does not contain any confidential, offensive, insulting, or threatening material.
Camera alignment. Object is in the scene center.
Object preview
- Downloads last month
- 115