What is FPN and why is it needed?

FPN (Feature Pyramid Network) is a neural network architecture used in computer vision for processing images at different scales. It constructs a feature pyramid, allowing for effective detection of objects of various sizes.

Why FPN is needed:

• Provides a multi-level feature representation, combining high-level semantics with low-level details.
• Improves the detection quality of small and large objects.
• Used in modern detection algorithms, such as Faster R-CNN, Mask R-CNN.

Principle of operation: FPN takes features from different layers of a base network (e.g., ResNet), enhances them, and combines them to create a set of feature maps with different resolutions, which are then used for detection.