Fast3r ViT Large 512

🚀 Fast3R - Towards 3D Reconstruction of 1000+ Images in One Forward Pass

Fast3R is a cutting - edge model aiming to achieve 3D reconstruction of over 1000 images in a single forward pass, which is presented at CVPR 2025.

🚀 Quick Start

📦 Installation

To use Fast3R in your own project, you need to follow the instructions from the Fast3R GitHub repo to install the necessary dependencies.

💻 Usage Examples

Basic Usage

from fast3r.models.fast3r import Fast3R
from fast3r.models.multiview_dust3r_module import MultiViewDUSt3RLitModule

# Load the model from Hugging Face
model = Fast3R.from_pretrained("jedyang97/Fast3R_ViT_Large_512")
model = model.to("cuda")

# [Optional] Create a lightweight lightning module wrapper for the model.
# This provides functions to estimate camera poses, evaluate 3D reconstruction, etc.
# See fast3r/viz/demo.py for an example.
lit_module = MultiViewDUSt3RLitModule.load_for_inference(model)

# Set model to evaluation mode
model.eval()
lit_module.eval()

📚 Documentation

📄 Citation

@InProceedings{Yang_2025_Fast3R,
    title={Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass},
    author={Jianing Yang and Alexander Sax and Kevin J. Liang and Mikael Henaff and Hao Tang and Ang Cao and Joyce Chai and Franziska Meier and Matt Feiszli},
    booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month={June},
    year={2025},
}

📄 License

The code and models are licensed under the FAIR NC Research License.