Add model card for QuerySplat
#2
by nielsr HF Staff - opened
README.md
CHANGED
|
@@ -1,3 +1,52 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
pipeline_tag: image-to-3d
|
| 4 |
---
|
| 5 |
+
|
| 6 |
+
# QuerySplat: Decoupling Geometry and Appearance Representations in 3DGS Prediction
|
| 7 |
+
|
| 8 |
+
[Project Page](https://inspatio.github.io/querysplat) | [Paper](https://huggingface.co/papers/2608.01186) | [GitHub](https://github.com/inspatio/QuerySplat)
|
| 9 |
+
|
| 10 |
+
QuerySplat is a feed-forward 3D Gaussian Splatting (3DGS) framework driven by geometric priors and explicit appearance decoupling. Specifically, it uses a dual-branch query-based decoder where the geometry branch leverages a pretrained Vision Geometric Model for spatial understanding, while the appearance branch recovers high-frequency details through a dedicated pathway separated from geometric attribute regression.
|
| 11 |
+
|
| 12 |
+
## Installation
|
| 13 |
+
|
| 14 |
+
```bash
|
| 15 |
+
git clone https://github.com/inspatio/QuerySplat.git
|
| 16 |
+
cd QuerySplat
|
| 17 |
+
|
| 18 |
+
conda create -n querysplat python=3.12 -y
|
| 19 |
+
conda activate querysplat
|
| 20 |
+
|
| 21 |
+
# Tested configuration: PyTorch 2.11.0 + CUDA 12.8.
|
| 22 |
+
python -m pip install torch==2.11.0 torchvision==0.26.0 \
|
| 23 |
+
--index-url https://download.pytorch.org/whl/cu128
|
| 24 |
+
python -m pip install --no-build-isolation -r requirements.txt
|
| 25 |
+
python -m pip install -U huggingface_hub
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## Inference
|
| 29 |
+
|
| 30 |
+
Place any number of images from one scene in `--input_folder`. Run inference with Test-Time Optimization (TTO):
|
| 31 |
+
|
| 32 |
+
```bash
|
| 33 |
+
python -m scripts.infer \
|
| 34 |
+
--config checkpoints/querysplat_vggto_1B_512_8192.yaml \
|
| 35 |
+
--checkpoint checkpoints/querysplat_vggto_1B_512_8192.safetensors \
|
| 36 |
+
--input_folder data/my_scene \
|
| 37 |
+
--output_dir outputs/my_scene \
|
| 38 |
+
--use_tto
|
| 39 |
+
```
|
| 40 |
+
|
| 41 |
+
Omit `--use_tto` to run the feed-forward model without test-time optimization.
|
| 42 |
+
|
| 43 |
+
## Citation
|
| 44 |
+
|
| 45 |
+
```bibtex
|
| 46 |
+
@article{li2026querysplat,
|
| 47 |
+
title={QuerySplat: Decoupling Geometry and Appearance Representations in 3DGS Prediction},
|
| 48 |
+
author={Li, Yinglong and Shen, Donghui and Zhang, Xiaoyu and Ye, Zhichao and Wu, Hongyu and Hao, Aimin and Zhang, Guofeng and Liu, Haomin},
|
| 49 |
+
journal={arXiv preprint arXiv:2608.01186},
|
| 50 |
+
year={2026}
|
| 51 |
+
}
|
| 52 |
+
```
|