Botao Ye · Boqi Chen · Haofei Xu · Daniel Barath · Marc Pollefeys
YoNoSplat reconstructs 3D Gaussian splats directly from unposed and uncalibrated images, while flexibly leveraging ground-truth camera poses or intrinsics when available.
## TODO - [x] Release code and pretrained models - [ ] Release high-resolution models - [ ] Dynamic dataloaders for training on more datasets - [ ] Release models trained on a mixture of more datasets ## Installation Our code requires Python 3.10+ and is developed with PyTorch 2.1.2 and CUDA 11.8, but it should work with higher PyTorch/CUDA versions as well. 1. Clone YoNoSplat. ```bash git clone https://github.com/cvg/YoNoSplat cd YoNoSplat ``` 2. Create the environment (example using conda). ```bash conda create -y -n yonosplat python=3.10 conda activate yonosplat pip install torch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 --index-url https://download.pytorch.org/whl/cu118 pip install -r requirements.txt ``` ## Pre-trained Checkpoints ## Pre-trained Checkpoints Our models are hosted on [Hugging Face](https://huggingface.co/botaoye/YoNoSplat) 🤗 | Model name | Training resolutions | Training data | |:-------------------------------------------------------------------------------------------------:|:--------------------:|:-------------:| | [re10k.ckpt]( https://huggingface.co/botaoye/YoNoSplat/resolve/main/re10k_224x224_ctx2to32.ckpt) | 224x224 | re10k | | [dl3dv.ckpt]( https://huggingface.co/botaoye/YoNoSplat/resolve/main/dl3dv_224x224_ctx2to32.ckpt ) | 224x224 | dl3dv | Download the checkpoints and place them in the `pretrained_weights/` directory. ## Camera Conventions The camera system follows the [pixelSplat](https://github.com/dcharatan/pixelsplat) convention: - **Intrinsics**: Normalized camera intrinsic matrices (first row divided by image width, second row divided by image height) - **Extrinsics**: OpenCV-style camera-to-world matrices (+X right, +Y down, +Z forward into the scene) ## Datasets Please refer to [DATASETS.md](DATASETS.md) for dataset preparation instructions. ## Running the Code ### Evaluation Please refer to [EVALUATION.md](EVALUATION.md) for detailed evaluation commands, including novel view synthesis, pose estimation, and metrics calculation. ### Training First download the [Pi3](https://huggingface.co/yyfz233/Pi3/resolve/main/model.safetensors) pretrained model and save it as `./pretrained_weights/pi3.safetensors` directory. #### Train on RealEstate10K (multi-view, 2-32 input views) ```bash python -m src.main \ +experiment=yono_re10k \ trainer.num_nodes=8 \ wandb.mode=online \ wandb.name=re10k_ctx2to32 \ optimizer.lr=1e-4 \ data_loader.train.batch_size=1 \ checkpointing.save_weights_only=false \ dataset.re10k.view_sampler.num_context_views=[2,32] ``` #### Train on DL3DV (multi-view, 2-32 input views) ```bash python -m src.main \ +experiment=yono_dl3dv \ trainer.num_nodes=8 \ wandb.mode=online \ wandb.name=dl3dv_ctx2to32 \ optimizer.lr=1e-4 \ data_loader.train.batch_size=1 \ checkpointing.save_weights_only=false \ dataset.dl3dv.view_sampler.num_context_views=[2,32] ``` You can adjust the batch size and number of GPUs/nodes to fit your hardware. Note that changing the total batch size may require adjusting the learning rate to maintain performance. ## Acknowledgements This project builds upon several excellent repositories: [NoPoSplat](https://github.com/cvg/NoPoSplat), [Pi3](https://github.com/yyfz/Pi3), [pixelSplat](https://github.com/dcharatan/pixelsplat). ## Citation If you find this work useful in your research, please cite: ```bibtex @inproceedings{ye2026yonosplat, title = {YoNoSplat: You Only Need One Model for Feedforward 3D Gaussian Splatting}, author = {Ye, Botao and Chen, Boqi and Xu, Haofei and Barath, Daniel and Pollefeys, Marc}, booktitle = {International Conference on Learning Representations (ICLR)}, year = {2026} } ```