What is Feature Pyramid Network (FPN)?
Updated May 15, 2026
Short answer
FPN builds multi-scale feature representations for object detection.
Deep explanation
FPN combines bottom-up feature extraction with top-down pathway and lateral connections. It allows detection of objects at different scales by merging high-resolution spatial features with deep semantic features.
Real-world example
Used in RetinaNet for detecting small and large objects in images.
Common mistakes
- Using only single-scale feature maps for detection tasks.
Follow-up questions
- Why are multi-scale features important?
- How does FPN improve small object detection?