Onboard image quality classification — prioritize high‑value frames

CubeSats are constrained by storage, processing, and slow communication. This UI mirrors the development pipeline—statistics‑driven preprocessing, a pruned CNN, and TFLite quantization—to rank images for downlink.

Drag & Drop or browse up to 10 images.

Client computes blur/noise metrics; server returns class + confidence.

Items: 0 • Tip: use filters to cluster by class