Computer Vision Interview Questions, issue 9, Jan 10, 2026

The Tiny Object Trap

Senior Computer Vision Engineer interview at Amazon Fulfillment Technologies & Robotics, and the interviewer asks:

β€œWe need to detect tiny, 3mm micro-fractures on a fast-moving assembly line. You suggested π…πšπ¬π­πžπ« 𝐑-𝐂𝐍𝐍 over π˜πŽπ‹πŽ. Why does the π‘πžπ π’π¨π§ 𝐏𝐫𝐨𝐩𝐨𝐬𝐚π₯ 𝐍𝐞𝐭𝐰𝐨𝐫𝐀 (𝐑𝐏𝐍) specifically help with small objects, even though it kills our inference speed?”

Don’t say: β€œBecause two-stage detectors are generally more accurate than single-stage detectors.’”

Why Faster R-CNN still beats YOLO when defects are smaller than your receptive field.

The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.

Read it on Substack

Get the next one

Free on Substack. Unsubscribe in one click.