Neural Implicit Isotropic Volume Reconstruction

Getting Started

conda create -n niv python=3.9
conda activate niv
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -c nvidia
pip install igneous-pipeline pytorch-ignite tqdm wandb

Start and Reconnect to Training Job using tmux

tmux new -s my_training_session
python train.py -opt options/train/train_iso_em.yml

Detach from the session using Ctrl+B D.

Reconnect to the session using

tmux attach -t my_training_session

List existing tmux sessions

tmux list-sessions

Delete tmux session

tmux kill-session -t session-name

Data Access

Requires gsutil command-line utility installed. See instructions here.

Download public training data from GCS

cd neural-volumes
gsutil cp gs://neural-implicit-volumes/datasets/hemibrain-volume-denoised-large.zip ./data
cd data
unzip hemibrain-volume-denoised-large.zip

Convert reconstructed volume to NG precomputed and upload to GCS

igneous image create ./DATA.npy ./PRECOMPUTED_FOLDER --compress none
gsutil -m cp -r ./PRECOMPUTED_FOLDER/  gs://neural-implicit-volumes/NAME/

References

We used the code from following repositories: NVP, LIIF, VINR.