Find weak samples in your TAO image classifier.
Identifies weak NVIDIA TAO VCN Classify samples, selects a decision threshold, and produces a ranked augmentation queue.
When to use it
Use when analyzing VCN classification failures, selecting augmentation targets, or auditing PASS/NO_PASS boundary cases.
Give it inference results, a training configuration, and dataset images; it writes a ranked augmentation queue, a gap-analysis report, and spot-check images.
What you provide
Requires Docker to inspect, pull, and run the data-services container.
Requires the NVIDIA container toolkit for the GPU-enabled container workflow.
This skill
Claude Code session transcript
Copies your agent session transcript
TAO data-services
Downloads the pinned TAO container