Skip to content

Getting Started

Follow these steps to set up the Cervical Cytology AI project on your local machine.

1. Prerequisites

Before starting, ensure you have the following installed:

  • Python 3.9+ (preferably 3.10 to 3.13)
  • Git
  • A CUDA-compatible GPU (highly recommended for faster training and inference)

2. Installation

Clone the repository and install the required dependencies:

git clone https://github.com/PritamTheCoder/cervical_cytology_ai.git
cd cervical_cytology_ai
pip install -r requirements.txt

Note: For PDF report generation, we explicitly use reportlab==4.1.0.

3. Dataset Setup

This project uses the SIPaKMeD Database to train the cell classifier.

  1. Download the dataset.
  2. Extract the dataset into the data/ directory.
  3. Ensure the structure uses class-wise folders (no pre-split Train/Test directories inside the main folder):
data/
  SIPAKMED/
    im_Dyskeratotic/
    im_Koilocytotic/
    im_Metaplastic/
    im_Parabasal/
    im_Superficial_Intermediate/

4. Downloading Model Weights

The classification portion requires trained MobileViT-S weights.

  • Option A: Use pre-trained weights. Download the checkpoint and place it in the weights/ folder. The expected file name is generally mobilevit_s_sipakmed_stain_normalized.pth (or similar, depending on your training run). Check out our HuggingFace repository for pre-trained weights if available.
  • Option B: Train your own weights from scratch by running the training scripts from the Usage Guide.

Next Steps

Once the environment is installed and data is available, move on to the Usage Guide to run the API, start the training pipeline, or generate clinical reports.