U-Net Segmentation Module

Train and run U-Net segmentation models on biomedical images. This module provides a complete pipeline from data upload through training to inference.

The U-Net Segmentation module provides a 4-step workflow for training deep learning models to segment biomedical images.

U-Net is a convolutional neural network architecture designed specifically for biomedical image segmentation. It uses an encoder-decoder structure with skip connections that preserve spatial information, making it particularly effective at segmenting objects in microscopy images.

Step 1: Upload your training images and corresponding annotation masks, or choose to use a pretrained model if you already have one.

Step 2: Configure training parameters like patch size, batch size, learning rate, and number of epochs. These settings control how the model learns from your data.

Step 3: Monitor training progress with real-time loss charts and Dice score metrics. Watch for convergence and potential overfitting.

Step 4: Run inference on new images using your trained model to generate segmentation masks.