What is U-Net architecture and how does it work in segmentation?
Updated May 15, 2026
Short answer
U-Net is a CNN architecture designed for precise image segmentation using encoder-decoder structure.
Deep explanation
U-Net consists of a contracting path (encoder) to capture context and an expansive path (decoder) to enable precise localization. Skip connections concatenate feature maps from encoder to decoder to preserve spatial detail lost during downsampling.
Real-world example
Used in biomedical image segmentation like detecting cells in microscopy images.
Common mistakes
- Ignoring skip connections which leads to poor spatial accuracy.
Follow-up questions
- Why is U-Net symmetric?
- What loss functions are used in segmentation?