Fundamentally, this project is a specialized computer vision toolkit designed to tackle the problem of cell tracking and segmentation in 2D microscope images. It serves as an adaptation of the DEGAST3D framework, retargeted and optimized specifically for two-dimensional data to maintain lineage and trace cellular division over time.
- Objective: To take raw 2D microscopic image sequences and output heavily analyzed tracks demonstrating where a cell moves, when it divides, and its exact contours.
- Workflow:
- Automatically loads and formats consecutive 2D microscopic datasets.
- Applies cellpose-based segmentation techniques adapted for 2D.
- Runs graph matching/graph sim neural network layers to track the coordinates and instances frame-by-frame.
- Flags cellular division points recursively throughout the tracking steps.
Inference_Tracking_Celldiv.py: Main execution script binding segmentation and the tracking logic together.cellpose3/&cellpose4/: Underlying segmentation engines.graph_sim/: Graph matching simulation sub-module acting as the mathematical backend for association across frames.checkpoints/: Model weights mapping PyTorch state dictionaries forcell_div,graph_pair, andtracking.
- Create a Conda environment with the required dependencies:
conda env create -f environment.yml - Activate the environment:
conda activate cell-lab
Run the tracking inference pipeline on your dataset:
python Inference_Tracking_Celldiv.py --data_path ./data/test_plant/microscopic_images/