Multi-Scale Feature Representations

U-Net

CNN/U-Net

PANet

adds an extra bottom-up path aggregation network on top of FPN

STDL

proposes a scale-transfer module to exploit cross-scale features

M2det

proposes a U-shape module to fuse multi-scale features

G-FRNet

introduces gate units for controlling information flow across features

NAS-FPN

autoML/NAS-FPN leverages neural architecture search to automatically design feature network topology.