TCT_data

A large annotated cervical cytology images dataset for AI models to aid cervical cancer screening. scientific data

Journal Link

Installation

Once you clone the repo, please run the following command to create the conda environment.

$ conda env create --file environment.yaml

Usage

Directory description:

├─ network             // directory of detection networks
├─ netdetr             // directory of detr network
├─ netsparse           // directory of sparse rcnn network
├─ netyolo             // directory of yolo network
├─ tool                // directory of tool codes

├─ datasets.py         // dataset code
├─ train.py            // main code for model training
├─ trainer.py          // code for training utils
├─ launch.sh           // train.py launcher
├─ _utils.py           // other utils code

Run the following code for model training:

$ bash launch.sh

This will initiate the script train.py for 5-fold cross-validation model training. Note that the csv file paths need to be changed according to the actual situation.