cyber2a / cyber2a-course

Online materials for the Cyber2A course on AI for Arctic research
https://cyber2a.github.io/cyber2a-course/
Apache License 2.0
0 stars 1 forks source link

Lesson - U-Net for semantic segmentation #10

Closed carmengg closed 7 months ago

carmengg commented 12 months ago

Deep dive into U-Net for semantic segmentation

Goal

To provide participants with a deep understanding of the U-Net architecture, its relevance to semantic segmentation tasks, and its application for RTS mapping.

Breakdown

  1. Introduction to Semantic Segmentation
    • Definition and significance of semantic segmentation
    • Differences between classification, object detection, and segmentation
    • Relevance to RTS mapping
  2. Overview of U-Net Architecture
    • Historical context: Why and where was U-Net developed?
    • Key features of U-Net: Symmetry, skip connections, etc.
    • Visual representation of U-Net's architecture
  3. Essentials of U-Net Components
    • Contracting path and its role in feature extraction
    • Bottleneck: Capturing the context
    • Expansive path: Localizing features using skip connections
  4. Introduction to Other Semantic Segmentation Models
    • FCN (Fully Convolutional Network): The pioneer in end-to-end segmentation
    • SegNet: Architecture with encoder-decoder structure
    • ...
  5. Case Study: U-Net for RTS Mapping
    • Walkthrough of a real-world application of U-Net for RTS mapping
    • Visualization of segmentation results on RTS data