license: mit
This is 125,160 (out of 800,000) 3D models from AllenAI Objaverse 1.0.
These models have had their materials & normals removed, have been normalised in scale and origin recentered, and are all triangulated.
As a result each model is much smaller in size, every model is uniform and faster to process.
This dataset is great for training networks on shapes of 3D objects.
My suggested use of this dataset is to raytrace a point-cloud of 𝑥 density from the triangle faced meshes, this can then be used to train a neural network to produce point-cloud shapes which can then be re-meshed.
The filenames will match up with the Allen-AI Objaverse Metadata.
If you want an even smaller dataset then Plyverse-1.0 is the same pre-processing on 10,000 3D models and includes the original vertex normals.
Possible licenses of these models, as quoted from AllenAI: