Results for “vertical-crop”

Extract false-positive and false-negative gaps from VLM binary-classification-question predictions by comparing model responses against ground truth, producing a structured JSONL file and summary report for downstream root-cause analysis.

Crops and reframes event photos to improve composition while preserving the original aspect ratio, with QA contact sheets and delivery folders.