This is the official code for 360° 3D Photos from a Single 360° Input Image
Manuel Rey Area,
Christian Richardt
University of Bath, Meta Reality Labs
IEEE VR 2025
After cloning repo, update submodules
git submodule update --init --recursive
We recomment running this code inside a coda environment
conda env create -f environment.yml
conda activate 3603dphoto
pip install submodules/diff-gaussian-rasterization
pip install submodules/simple-knn
Follow the installation steps for 360monodepth. Add the following line: return estimated_depthmap, erp_rgb_image_data here
The intialization stage can be run as following:
python train.py -s path/to/input_rgb.jpg -m exp1 --random_background --eval --sh_degree 0 --iterations 7000
Note: if ground-truth depth is available, provide the file with --ground-truth path/to/gt_depth.dpt
Next, the scene completion stage is run as:
python pipeline.py --eval -m exp1 --start_iteration 7000 --fill_strategy SDInp --sh_degree 0 --random_background --xp_name virtual
To render a lemniscate camera trajectory:
python render.py --eval -m exp1/virtual --lradius 0.5 --pathc lemniscate
Finally, to evaluate the renderings quality:
python metrics.py -m exp1/virtual