Home▸AI & Machine Learning▸2D Feature Pyramid Network: Scale Assignment, Top-Down Merging and Noise Averaging for Small Objects

🔺 2D Feature Pyramid Network: Scale Assignment, Top-Down Merging and Noise Averaging for Small Objects

Maps of a feature pyramid side by side: backbone C2 to C5, top-down P2 to P5, the level the FPN rule picks for a target square, the shared head response, noise averaging by stride and anchor IoU, with five independent numerical checks.

AI & Machine Learning2DAdvanced60 FPS📱 Mobile-adapted🟠 Blender⇄ 3D version
2D Feature Pyramid Network: Scale Assignment, Top-Down Merging and Noise Averaging for Small Objects ↗ Open standalone
⚙ Under the hood

Maps of a feature pyramid side by side: backbone C2 to C5, top-down P2 to P5, the level the FPN rule picks for a target square, the shared head response, noise averaging by stride and anchor IoU, with five independent numerical checks.

Feature Pyramid NetworkObject DetectionMulti-Scale FeaturesAverage PoolingTop-Down PathwayAnchor IoUSignal-to-Noise RatioComputer Vision

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install

What did you find?

Add reproduction steps (optional)