Reading a detector's module graph

A top-to-bottom map of a real detection network. Pairing execution order with resolution is what makes an FPN or PAN legible: the backbone walks down and right as it downsamples, and the top-down path climbs back left.

Try this

Follow one feature map from the backbone to the head and count how many times its resolution changes on the way.

backbone neck · PAN heads

Each row is one step of the forward pass, read downward. Columns are the size of the feature map, shrinking left to right as the network downsamples. Click a group to open it.

Where this lab comes from

YOLO-NAS Strikes Back: Still the Fastest Detector on Hardware You Already Own

Quantized to INT8, YOLO-NAS-S runs at 270 FPS on a laptop CPU and 1146 FPS on a mid-range laptop GPU, and it drops straight into Frigate. The model aged well; its tooling did not. Here is a clean reimplementation, measured across CPU, Intel iGPU and NVIDIA dGPU, and checked bit-for-bit against the original.