Below are some models I've put together over the course of my research. In contrast to scanned data, which often provides only the surface geometry, the idea here is to provide a few models that include all that other stuff like textures, control cages, etc. Note, however, that many of these models are not ideal for testing the robustness of some algorithms since most of them are noise-free and perfectly manifold—compare with the models available from the Stanford 3D Scanning Repository and the AIM@Shape Project.
Wherever possible I have released data into the public domain—the terms of this dedication are spelled out in a CC0 1.0 Universal (CC0 1.0) Public Domain Dedication. Basically this dedication says that you can do whatever you like with the data without asking for permission or acknowledging the source (although acknowledgement is certainly appreciated!). I specifically request that the editors of ACM Transactions on Graphics do not require authors to include copyright information for this material in figure captions or videos.
Several people have asked me for the model of this ogre, who starred in our ray tracing paper. Since there are few models available to researchers which include anything beyond a triangle mesh, I've also included the following data:
Building inspired by the San Diego convention center (originally designed by Arthur Erickson). The archive contains:
This dataset is a growing archive of 3D models created by students in the intro computer graphics class at Carnegie Mellon University. Each mesh was created by a different student, using their own implementation of a polygonal mesh editor / subdivision modeler, within the pedagogical software framework Scotty3D. Models are not guaranteed to be closed, manifold, or free of self-intersections; many models have inconsistent orientation across different components. We are hopeful that this data will be useful to algorithm and application developers interested in serving novice/non-professional users. In addition to surface geometry, many of these models include material attributes and named parts. Some models may contain additional scene data (e.g., lights or cameras).
Cloth motion capture data from our paper Capturing and Animating Occluded Cloth, stored as sequences of OBJs (with texture coordinates).
A spotted animal homeomorphic to a sphere (audio courtesy Clarus). Comes with Catmull-Clark control mesh, quadrangulation, triangulation, vector texture, and bitmap texture.
An animal that would never survive in the wild. Comes with Catmull-Clark control mesh, quadrangulation, triangulation, vector texture, and bitmap texture.
A friendly creature from our paper on stripe patterns. Comes with Catmull-Clark control mesh, quadrangulation, triangulation, vector texture, and bitmap texture.
Snakeboard model from our paper on vehicle animation. Broken down into individual articulated pieces.
High-resolution mesh of a sleeve from our paper Capturing and Animating Occluded Cloth. Single OBJ with texture coordinates; 18892 vertices, 37580 triangles.
As in, yeah, right, we can come up with a good texture atlas for this... The model has the following notable properties:
An abstract city. The archive contains:
The developable oloid surface at various resolutions, as both a quad and triangle mesh. Also includes code to generate the oloid at any target resolution.
Mechanically, the entire surface of the oloid comes in contact with the ground as it rolls along a linear trajectory. Here's a rigid body simulation of the model, courtesy Breannan Smith: Movie (QuickTime, 588KB)
A scan acquired in the City of Bridges, at CMU's Panoptic Studio. The archive includes data from several stages of the reconstruction process, starting with the raw point cloud all the way to a closed, remeshed surface with nice triangles. The final surface has genus 3, since it includes part of the ground. Many thanks to Tomas Simon, Tim Godisart, and Stelian Coros.
The archive contains:
A polyhedral torus where every pair of vertices is connected by an edge. This surface was originally described by Ákos Császár in the paper "A Polyhedron without Diagonals," Acta Sci. Math (1949). While there are several versions of this surface floating around on the web, none of them seem to agree with the description given in the paper (and some are not even embedded!). Part of the issue, perhaps, is that the original paper seems to be available only as a noisy, low-resolution scan. The archive below contains a carefully-checked reproduction of the original paper, as well as a mesh where vertex coordinates and indices correspond exactly to the original values. Note however that the paper does not describe an oriented polyhedron—the archive hence contains both the original version and an oriented version.
The archive contains:
This archive contains meshes of all of the regular and semi-regular tilings of the plane. Here a "regular" tiling is one consisting of only a single regular polygon. A "semi-regular" tiling may include more than one type of polygon, but every vertex will look identical.
The archive includes:
Each mesh contains roughly two thousand faces. The source can be used to generate tilings with more faces.
Though this mesh looks like it has a lot of bumps and wrinkles, every vertex is intrinsically flat: the angles around each around each interior vertex sum to exactly 2π, which means the mesh can be flattened perfectly into the plane without any distortion of lengths, areas, or interior angles. (Such a flattening is stored in the texture coordinates.) The mesh may be useful for testing the isometry invariance of geometry processing algorithms, i.e., the degree to which their results are unchanged by motions of the vertex positions that do not change edge lengths, areas, etc.
Naively, one sometimes calls such a mesh “discretely developable,” though this naive interpretation of developability neglects other important features of smooth developable surfaces---such as the presence of straight lines passing through each point of the surface. This phenomenon is studied in our paper on the developability of triangle meshes.
The mesh is stored as a Wavefront OBJ file with vertex and texture coordinates.
The meshes in this archive describe a polyhedral version of the Klein bottle, as well as a printable pattern for constructing this polyhedron out of paper, cardboard, or other flat material. Since the Klein bottle intersects itself, it is impossible to form the polyhedron from a single flat sheet. Instead, the instructions suggest to build the "box" and "J" pieces separately, insert the J in the box, then fill in the region of intersection with a final square of material.
This archive contains a bust of the Egyptian queen Nefertiti, comprised of about 2M triangles. The mesh was scanned by Nora Al-Badri and Jan Nikolai Nelles from the Nefertiti bust, which was created in 1345 BC by Thutmose. You can read more about the project here.
