Deep learning for image analysis block within the MAPI3 MSc: from perceptrons to deep and convolutional neural networks, with a practical focus on optimization and training.

Deep Learning for Image Analysis

• Perceptron: core building block and learning principle
• Deep neural networks: representation learning for vision tasks
• Convolutional Neural Networks (CNNs): convolutional layers and inductive biases for images
• Optimization for neural networks: training objectives and gradient-based learning workflows

Additional Image Processing & Analysis Topics

• Filtering & mathematical morphology; image segmentation: (thresholding, region growing, PDE- or Bayesian-based modeling)
• Image registration: deformation and similarity models, deformation gradients, diffeomorphic methods, average image computation
• Color/contrast processing & compression: RGB/HSV spaces, optimal transport; spectral methods, SVD, compressed sensing

Hands-on (Python Labs)

• Practical sessions to connect models to real image problems, and to discuss strengths/limits in applied settings