New version of MVTec HALCON focuses on speed and performance of AI and rule-based methods

“The focus of the new HALCON release is speed. Our goal is for machine vision applications across all industrial sectors, including demanding automation scenarios, to operate not only robustly and precisely but also at exceptional speed. For HALCON 26.05, we have improved both the speed and performance of deep learning methods and rule-based approaches,” explains Jan Gärtner, Product Manager HALCON at MVTec.

For example, the new version includes a newly integrated rectification function in HALCON. This enables Data Matrix codes on curved or deformed surfaces to be read quickly and reliably. In addition, automated contour optimization is now available for shape matching. The optimization removes unstable or misleading contours based on sample images, making matching faster, more stable, and more accurate. MVTec is also expanding functionality in the area of AI-based object detection. The new generation of deep learning-based object detection in HALCON 26.05 enables significantly faster inference while maintaining high detection accuracy. Detection remains reliable even for small objects and strongly varying object sizes. Integrated data augmentation techniques also increase robustness against lighting changes, rotations, and occlusions. 
As part of HALCON 26.05, a new preview version of MVTec's integrated development environment, HDevelopEVO, is also available. With the new release, it is now also possible to integrate scripts created in HDevelopEVO into your own applications using the .NET interface. In addition, multimodal LLM support has been expanded to include visual prompting. Developers can now incorporate image data directly into prompts for the AI Assistant.

Automatic contour optimization for Shape Matching: Reflections, shadows, and texture often introduce unstable contours that reduce matching reliability and require manual cleanup. The new automatic contour optimization in HALCON 26.05 removes these unreliable contours based on sample images, making matching faster, more stable, and more accurate — especially for reflective or textured objects in demanding automation scenarios.

Data Matrix rectification: Curved or deformed surfaces distort code geometry and reduce reading reliability. The new rectification capability in HALCON 26.05 compensates for these distortions before decoding and integrates optionally into existing workflows, with typical applications including cylindrical components, curved packaging, and flexible materials.

Enhanced data augmentation: A new operator-based approach in HALCON 26.05 replaces procedure-based augmentation and integrates directly into deep-learning pipelines. By defining flexible augmentation pipelines programmatically, developers improve model robustness and generalization and reduce reliance on large training datasets.

Next-generation AI object detection: A new generation of deep-learning-based object detection in HALCON 26.05 delivers up to 5x faster inference while maintaining high detection accuracy. It performs reliably even for small objects and varying object sizes, and integrated data augmentation increases robustness against illumination changes, rotation, and occlusion.

Downloads

• Press Release: MVTec New HALCON version 26.05
• Image: Jan Gärtner, Product Manager HALCON
• Image: Faster interference for deep learning based object detection

Legal Notice
All rights reserved, Copyright ©MVTec Software GmbH. All image and text material may be used solely for press coverage for free. Any kind of commercial usage of the material beside press coverage is prohibited.

We would be pleased to receive a voucher copy.