PanNuke Semantic Segmentation Project

This repository contains implementations of deep learning models for nuclei instance segmentation and classification using the PanNuke dataset. The project explores different architectural approaches including U-Net and DeepLabV3+ with ResNet50 for semantic segmentation tasks in histopathological images.

Dataset

The PanNuke dataset consists of:

• 189,744 labeled nuclei with instance segmentation masks
• 7,901 images (256×256 pixels)
• 19 tissue types
• 5 cell categories (Neoplastic, Inflammatory, Connective, Dead, Epithelial)
• Images captured at x40 magnification (0.25 µm/pixel resolution)

Implementation

The project includes three main implementations:

1. Basic U-Net architecture
2. ResNet50 with U-Net
3. ResNet50 with DeepLabV3+

Each implementation is available in separate Jupyter notebooks in the notebooks/ directory.

Methods

The implementations utilize state-of-the-art deep learning architectures for semantic segmentation:

• U-Net: Convolutional network architecture specifically designed for biomedical image segmentation
• ResNet50: Deep residual network used as a backbone for feature extraction
• DeepLabV3+: Advanced semantic segmentation architecture incorporating atrous convolutions

Reference Papers

• U-Net: Convolutional Networks for Biomedical Image Segmentation

License

This project is licensed under the terms included in the LICENSE file.

Author

Nikolaos Tsopanidis (aivc24022)