ghost-hunter

TODO: Create way of doing inference on multiple dm4 files

Project Workflow:

Training (only need to do once)

• Retrieve dm4 files
• Turn to png using scripts/dm4_to_png.py
• Label 500 images or so using labelling/binary_search_labelling
• setup yaml file with location of png files and select model settings
• train on labelled images using src/run.py
• Run inference using checkpoint using src/inference.py (real space segmentation will be output to images/)

Inference

• Retrieve dm4 files
• Turn to png using scripts/dm4_to_png.py
• setup yaml file with location of png files and select model settings
• Run inference using checkpoint using src/inference.py (real space segmentation will be output to images/)

Note

All the file imports have been setup so that they are relative to the top folder. So that when you reference files in the config files, always use folder paths relative to the top repo folder.

Usage

The config files are contained within the configs directory. You can control all the parameters of the model within this file. If you are running this model with your own data you will want to change the train_dir and data_dir directory. The project is also set up with wandb so and you can change the settings of this under the wandb section.

Within the jobs/ section you will find the pre-made bash scripts for running the experiments shown in the paper. But if you wish to run without them you can simply do so as following:

Training

python -m src.run "CONFIG_FILE.yaml" 

Inference

python -m src.inference "CONFIG_FILE.yaml"