Open model on one GPU reads 8 medicine boxes in 5 languages, flags duplicate ibuprofen in 4s

MaziyarPanahi · x · 2026-11-04

A demo showing a single open model on one GPU reading all 8 medicine box photos across 5 languages and outputting an English list in 4 seconds — correctly flagging that two boxes were the same ibuprofen, an easy-to-miss duplicate on a busy ward.

The author ran a consistency check (same answers over 3 runs) and noted the photos are public Wikimedia Commons images (CC0 / CC BY-SA 4.0), not one patient's bag.

Related event: Open-source model reads multilingual medicine boxes in 4 seconds on a single GPU(2 posts)→

Original post →

More from Apps

Apps channel →