CountPillar is a Python-based tool to generate synthetic datasets for counting and detecting pills. This repository creates synthetic images by adding pills to background images (e.g., a plate) to help develop and test machine learning models for pill counting and detection tasks.
To install CountPillar, you will need to create a conda environment and install the required dependencie
-
Create and activate a new conda environment:
conda create -n countpillar python=3.10 conda activate countpillar -
Install all the dependencies using:
pip install -r requirements.txt -
Install the CountPillar package using:
pip install -e .
To set up a development environment for CountPillar, you can use the following command:
pip install -e .["dev"]
This will install additional development dependencies and allow you to make changes to the CountPillar code.
This repository uses pre-commits for flake8, black, and isort.
Run the following command from the root of the repository:
pip install pre-commit
pre-commit installto ensure that all code is properly linted & formatted before committing.
To use CountPillar, you will first need to generate masks for the pill images in the ./data/pills/images directory. This can be done by running the following command:
mask_generator
This will create mask images for each pill image and save them in the ./data/pills/masks directory.
After obtaining all the masks, you can create a synthetic dataset by running the following command:
dataset_generator
The generated dataset will be stored in the ./dataset/synthetic directory, with the following structure:
dataset/
└── synthetic/
├── images/
└── labels/
The images folder contains the synthetic images of pills composed on the background. The number of pills in each image is indicated in the filename. For example, a file named 0_18.jpg implies that there are 18 pills in the image.
The labels folder contains bounding box annotations for each generated image, stored in the YOLO format. These annotations can be used for training and evaluating object detection models, specifically tailored for counting and detecting pills in images.