Views
No views yet
| File | Size | Description |
|---|---|---|
autolevel.pth | 107 MB | Geometry estimation model combining a pretrained SegNeXt-B backbone with a trained flow decoder and rigid filter. |
pytorch_lora_weights.safetensors | 171 MB | Fine-tuned panorama-generation LoRA weights for FLUX.1-Fill-dev. |
1git clone https://github.com/Orange-3DV-Team/Gimbal360.git
2cd Gimbal360
3
4conda create -n gimbal360 python=3.10 -y
5conda activate gimbal360
6pip install -r requirements.txt
7
8python inference.py \
9 --image example/example.png \
10 --prompt "a frozen, abandoned train car in a post-apocalyptic wasteland." \
11 --resolution 9601hf download Orange-3DV-Team/Gimbal360 --local-dir ./ckpt
2
3python inference.py \
4 --image example/example.png \
5 --prompt "a frozen, abandoned train car in a post-apocalyptic wasteland." \
6 --ckpt_dir ./ckpthuggingface_hub:1from huggingface_hub import hf_hub_download
2
3autolevel_path = hf_hub_download("Orange-3DV-Team/Gimbal360", "autolevel.pth")
4lora_path = hf_hub_download("Orange-3DV-Team/Gimbal360", "pytorch_lora_weights.safetensors")pipeline.load_lora_weights(lora_path).autolevel.pth) estimates the camera geometry.pytorch_lora_weights.safetensors) inpaints the full 360° panorama.1@article{lu2026gimbal360,
2 title={Gimbal360: Canonicalizing Planar Diffusion for Spherical Panorama Completion},
3 author={Lu, Yuqin and Liu, Haofeng and Zhou, Yang and Dai, Yihua and Li, Guiqing and He, Shengfeng and Liang, Jun},
4 journal={arXiv preprint arXiv:2603.23179},
5 year={2026}
6}