Download instructions:

Clone this repo to a directory of your choice. Download HLOC and KAIR submodules

git submodule update --init --recursive

Installation instructions:

1. If not yet installed, install a conda distribution for our enviroment. I recommend the conda-forge installer. 1a. If other conda distributions are used i'd still suggest setting conda-forge as a primary channel: conda config --add channels conda-forge conda config --set channel_priority strict
2. Create and activate a new environment, with python 3.10 and pip conda create -n computer_vision python=3.10 pip conda activate computer_vision You may use mamba instead of conda if you prefer.
3. If you have an Nvidia GPU, check your driver version; I'll use cuda version 11.8 in this example, but it should be the same, for others, as long as you replace 11.8 with your version.
4. Install pytorch 2.1.2 for your CUDA version (we use 2.1.2 for compatibility reasons, although it may work with newer versions). Check here on how to install with ROCM/cpu: conda install pytorch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 pytorch-cuda=11.8 -c pytorch -c nvidia
5. Install colmap 3.12 and pycolmap 3.12. If a version with your CUDA driver is available it will automatically download the GPU-enabled version. conda install colmap=3.12 pycolmap=3.12
6. Install the complete CUDA package of your version from the nvidia channel (skip if cpu only): conda install nvidia/label/cuda-11.8.0::cuda
7. Install the tiny-cuda-nn bindings (may skip if no nvidia or if some wrong happens, it is just a speedup), may take a while: pip install --upgrade-strategy only-if-needed ninja 'git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch'
8. Install the requirements for KAIR: cd KAIR pip install --upgrade-strategy only-if-needed -r requirement.txt cd ..
9. Install Hierarquical-Localization for Superpoint/SuperGlue/LoFTR: cd Hierarchical-Localization pip install --upgrade-strategy only-if-needed -e . cd ..
10. Install Nerfstudio: pip install --upgrade-strategy only-if-needed nerfstudio
11. You MAY run into some issues regarding numpy versions; check by running ns-train --help and if some errors regarding numpy come up, reinstall numpy to version 1.26.4 (you may also need to reinstall opencv): `pip install numpy==1.26.4 opencv-python'
12. Install ipympl for interactive matplotlib pip install ipympl

Post-installation instructions:

1. Download pretrained models from KAIR: python main_download_pretrained_models.py --models "others" --model_dir "model_zoo" python main_download_pretrained_models.py --models "DnCNN" --model_dir "model_zoo" python main_download_pretrained_models.py --models "SwinIR" --model_dir "model_zoo"