DeepOrientation: Deep Orientation Estimation of Macromolecules in Cryo-electron tomography

This repository is the implementation of DeepOrientation: Deep Orientation Estimation of Macromolecules in Cryo-electron tomography.

DeepOrt is a learning-based network for orientation estimation based on six degrees of freedom of the object (6DoF). The network architecture includes a multi-layer perceptron.

Data availability

Our data is available at: https://www.zib.de/ext-data/PolNet_Medium_Size_Dataset_4v4r_and_3j9i.zip

Training DeepOrt

In case you are running on a single GPU/CPU workstation, please simply run:

python train_slurm.py

In case you would like to run a slurm job, set the following paths in the submit_tf.sh file:

#SBATCH --error=<path/to/result/foldr/deeport_errors_%j.err>
# job output file
#SBATCH --output=<path/to/result/folder/deeport_output_%j.out>

Then using the following command, starts the training on the cluster. sbatch submit_tf.sh

In order to set training parameters, please use the config.py file in the main root.

The data folder should have the following structure:

data
 |____ <dataset name>
       |_____ npy
       |_____ res

The npy folder holds the npy files (euler, quaternions, 6dof), train and test data. While the res folder is where the results will be dumped.

citation: