MVTec made Deep Learning Tool 26.10 available on Oct. 30. The industrial-vision release imports images and labels in COCO or YOLO format, adds evaluation to its out-of-distribution detection workflow, and allows separately trained models to be evaluated. Deep OCR users can inspect individual character results and inference times. These are software workflow changes, not measured improvements in inspection accuracy.
GLOBAL IMAGING BRIEF
Analysis: Reusing labeled datasets can reduce migration work when comparing models, but production teams still need to validate label mapping, class drift, latency and target HALCON compatibility on their own images.
Original source
Original sources: MVTec's Oct. 30 announcement and version 26.10 release notes. No independent accuracy benchmark is implied.