Hi! I'm Niels from the community science team at Hugging Face.
This PR adds the `pipeline_tag: image-to-3d` to the model card metadata. This helps categorize the model correctly on the Hugging Face Hub and makes it easier for the community to discover your work in the 3D reconstruction and video processing space.
---
tags:
- computer-vision
- 3d-reconstruction
- video-processing
---
# LoGeR: Long-Context Geometric Reconstruction with Hybrid Memory
This repository contains the reimplemented model checkpoints for **LoGeR**.
LoGeR processes long video streams in chunks with a hybrid memory design to improve large-scale geometric reconstruction quality and consistency.
[](https://LoGeR-project.github.io/)
[](https://github.com/Junyi42/LoGeR)
[](https://arxiv.org/abs/2603.03269)
<p align="center">
<img src="https://loger-project.github.io/figs/fig1_teaser.png" alt="LoGeR Teaser" width="100%">
</p>
## Checkpoints
We provide two main pre-trained models. To use them, please clone the [corresponding GitHub repository](https://github.com/Junyi42/LoGeR) and place the downloaded `.pt` files in your local `ckpts…
