Views
No views yet
1git clone https://github.com/inspatio/QuerySplat.git
2cd QuerySplat
3
4conda create -n querysplat python=3.12 -y
5conda activate querysplat
6
7# Tested configuration: PyTorch 2.11.0 + CUDA 12.8.
8python -m pip install torch==2.11.0 torchvision==0.26.0 \
9 --index-url https://download.pytorch.org/whl/cu128
10python -m pip install --no-build-isolation -r requirements.txt
11python -m pip install -U huggingface_hub--input_folder. Run inference with Test-Time Optimization (TTO):1python -m scripts.infer \
2 --config checkpoints/querysplat_vggto_1B_512_8192.yaml \
3 --checkpoint checkpoints/querysplat_vggto_1B_512_8192.safetensors \
4 --input_folder data/my_scene \
5 --output_dir outputs/my_scene \
6 --use_tto--use_tto to run the feed-forward model without test-time optimization.1@article{li2026querysplat,
2 title={QuerySplat: Decoupling Geometry and Appearance Representations in 3DGS Prediction},
3 author={Li, Yinglong and Shen, Donghui and Zhang, Xiaoyu and Ye, Zhichao and Wu, Hongyu and Hao, Aimin and Zhang, Guofeng and Liu, Haomin},
4 journal={arXiv preprint arXiv:2608.01186},
5 year={2026}
6}