We recommend using Conda to install the necessary dependencies. To do so, run the following commands:
conda env create -f environment.yml
conda activate ppicker
To use ProPicker, you need the checkpoint of our pre-trained model, as well as the checkpoint of the TomoTwin model we used as prompt encoder:
-
You can download the ProPicker checkpoint here here
-
You can download the TomoTwin checkpoint by running
bash download_tomotwin_ckpt.sh
After downloading, place the files in the ProPicker directory. If you want to store them somewhere else, you have to adjust the paths in paths.py.
We provide an example for prompt-based picking in the TUTORIAL1 notebook, in which we pick ribosomes in the EMPIAR-10988 dataset.
An example for fine-tuning ProPicker on the EMPIAR-10988 dataset is provided in the TUTORIAL2 notebook.
Training is handled in the train.py script. All necessary parameters are set in train_cfg.py.
To download the training data, you can use the the datasets/download_train_data.sh script.
Note: The training data is large, so you might want to download it to a different location. To do this, you can modify the download_train_data.sh script. In this case, you also have to adjust the path to the training data in paths.py
This repository contains code copied and modified from the following projects:
All derived code is explicitly marked as such.