Test-time training 3D reconstruction

Researchers released TTT3R, an open-source 3D reconstruction method that treats reconstruction as test-time training. The project provides code, pretrained checkpoints, and inference demos for converting video or image sequences into 3D models.

ttt3r.mp4 - Clone TTT3R. git clone https://github.com/Inception3D/TTT3R.git cd TTT3R - Create the environment. conda create -n ttt3r python=3.11 cmake=3.14.0 conda activate ttt3r conda install pytorch torchvision pytorch-cuda=12.1 -c pytorch -c nvidia use the correct version of cuda for your system pip install -r requirements.txt issues with pytorch dataloader, see https://github.com/pytorch/pytorch/issues/99625 conda install 'llvm-openmp<16' for evaluation pip install evo pip install open3d - Compile the cuda kernels for RoPE as in CroCo v2 . cd src/croco/models/curope/ python setup.py build ext --inplace cd ../../../../ CUT3R provide checkpoints trained on 4-64 views: cut3r 512 dpt 4 64.pth https://drive.google.com/file/d/1Asz-ZB3FfpzZYwunhQvNPZEUA8XUNAYD/view?usp=drive link . To download the weights, run the following commands: cd src gdown --fuzzy https://drive.google.com/file/d/1Asz-ZB3FfpzZYwunhQvNPZEUA8XUNAYD/view?usp=drive link cd .. To run the inference demo, you can use the following command: input can be a folder or a video the following script will run inference with TTT3R and visualize the output with viser on port 8080 CUDA VISIBLE DEVICES=6 python demo.py --model path MODEL PATH --size 512 \ --seq path SEQ PATH --output dir OUT DIR --port 8080 \ --model update type ttt3r --frame interval 1 --reset interval 100 \ --downsample factor 1000 --vis threshold 5.0 Example: CUDA VISIBLE DEVICES=6 python demo.py --model path src/cut3r 512 dpt 4 64.pth --size 512 \ --seq path examples/westlake.mp4 --output dir tmp/taylor --port 8080 \ --model update type ttt3r --frame interval 1 --reset interval 100 \ --downsample factor 100 --vis threshold 6.0 CUDA VISIBLE DEVICES=6 python demo.py --model path src/cut3r 512 dpt 4 64.pth --size 512 \ --seq path examples/taylor.mp4 --output dir tmp/taylor --port 8080 \ --model update type ttt3r --frame interval 1 --reset interval 50 \ --downsample factor 100 --vis threshold 10.0 Output results will be saved to output dir . Please refer to the eval.md /Inception3D/TTT3R/blob/main/eval/eval.md for more details. Our code is based on the following awesome repositories: We thank the authors for releasing their code If you find our work useful, please cite: @article{chen2025ttt3r, title={TTT3R: 3D Reconstruction as Test-Time Training}, author={Chen, Xingyu and Chen, Yue and Xiu, Yuliang and Geiger, Andreas and Chen, Anpei}, journal={arXiv preprint arXiv:2509.26645}, year={2025} }