Last updated 08-13-2026
Category:
Reviews:
Join thousands of AI enthusiasts in the World of AI!
Farm3D
Farm3D is a research method for single-view 3D reconstruction of articulated animals, built at the University of Oxford and published at 3DV 2024. You feed it one photo of a horse, cow, or sheep and it returns a full 3D shape with texture in seconds, without training on real photographs. The project page documents the approach, shows reconstruction demos, and links to code and the Animodel benchmark dataset.
Most 3D generators need large labeled 3D datasets or multi-view captures. Farm3D trains entirely on synthetic views produced by Stable Diffusion, then uses the same diffusion model as a critic during learning. That lets it recover fine details like legs and ears on categories it never saw in real images, which is unusual for monocular reconstruction pipelines that rely on real-world supervision.
Researchers studying articulated 3D shape, computer vision engineers prototyping animal asset pipelines, and 3D artists exploring controllable synthesis will find the demos and open code useful. The method also supports relighting, texture swapping between same-category models, and skeletal animation once a shape is generated.
Reconstructs articulated 3D animal shapes from one input image in seconds
Trains without real photos by distilling virtual views from Stable Diffusion
Factorizes each instance into shape, albedo, diffuse and ambient lighting, viewpoint, and light direction
Supports relighting, texture swapping between same-category models, and skeletal animation on generated assets
Animodel benchmark ships textured meshes for horses, cows, and sheep with realistic articulated poses
Paper accepted at 3DV 2024; code and dataset published on GitHub under tomasjakab/animodel
Reconstructs fine articulated detail from a single photo without real-image training data.
Open paper, code, and Animodel benchmark are all linked from the project page.
Generated assets support relighting, texture swap, and animation in the published demos.
Research project, not a hosted app or API you can sign up for and run in a browser.
Focused on articulated animal categories such as horses, cows, and sheep rather than general objects.
Running the code requires setting up the GitHub repository and compute environment locally.
What does Farm3D reconstruct from a photo?
Farm3D rebuilds an articulated 3D animal from a single input image. The method outputs shape, texture, and pose controls, and the project page shows horses, cows, and sheep reconstructions with fine leg and ear detail.
Is Farm3D free to use?
Yes, Farm3D is free research software. The project page links to an arXiv paper, BibTeX entry, and GitHub repository with code and the Animodel benchmark at github.com/tomasjakab/animodel.
Does Farm3D need real training images?
No. Farm3D learns categories entirely from synthetic views prompted through Stable Diffusion. During training it also samples synthetic instance views that Stable Diffusion critiques to guide reconstruction quality.
What can you do after Farm3D generates a 3D asset?
Farm3D outputs controllable 3D assets you can relight, animate, or texture-swap with another model in the same category. The project page shows animation, relighting, and texture swapping demos on generated shapes.
What is the Animodel dataset in Farm3D?
Animodel is a 3D articulated animals dataset bundled with Farm3D for benchmarking single-view reconstruction. It includes professionally crafted textured meshes of horses, cows, and sheep in realistic articulated poses.
Where was Farm3D published?
Farm3D was developed at the University of Oxford and published at 3DV 2024. The arXiv preprint is titled Farm3D: Learning Articulated 3D Animals by Distilling 2D Diffusion, arXiv:2304.10535.