AI & Machine Learning

Tao Run DEFT AOI Cosmos3

The skill executes a disk‑backed DEFT AOI improvement loop for NVIDIA Cosmos Reason 3 and Cosmos3 models, evaluating the base model on Proxy and frozen Benchmark splits, mining real image pairs from Proxy gaps, and generating synthetic NG pairs with AnomalyGen to assemble per‑iteration training data for Cosmos Framework LoRA SFT. It trains and repeats the process on the chosen platform, delivering only OK or NG labels for each inspection. This workflow empowers engineers to refine Cosmos3 PCB inspection models with minimal annotation effort.

Publisher / source

NVIDIA Skills (Verified)

Upstream record

NVIDIA/skills

Version

Version 3

Published

2026-09-23

Publication review

Approved for publication

Content fingerprint

faa153ae3e516879705beeafe0720604c22ebfd7f639091239931e869940846d