Deep learning vision inspection software company Neurocle has announced the release of version 4.4 of its Auto Deep Learning-based inspection software. The latest update focuses on expanding the applicability of deep learning models across diverse industrial environments, while maximizing the speed and efficiency of the model development workflow.

Hardware environments have a direct impact on performance when it comes to developing deep learning models. As such, securing scalability in hardware and platform infrastructure enabling optimized model implementation across various systems has long stood as a key challenge. Furthermore, the automation and optimization of labeling and training stages, which typically require significant time and human resources, remain an area in need of continuous improvement.

Reflecting these demands, version 4.4 delivers enhanced scalability and productivity for users who develop and deploy models directly in manufacturing environments.

Support for NVIDIA RTX 50 Series – Optimizing AI Deep Learning Model Development and Inference in Next-Generation Hardware Environments

The first key update in version 4.4 is support for NVIDIA’s latest GPU lineup, the RTX 50 series. By ensuring compatibility with the new RTX 50 series at launch, Neurocle enables seamless model training and inference even in cutting-edge GPU environments.

This provides significantly improved computational performance, reducing the time required from the creation of deep learning inspection models to their deployment in the field. Furthermore, with a broader range of hardware infrastructure options now available, the flexibility and scalability to develop optimized models across diverse system environments have been further enhanced.

Enhanced AI Auto-Labeling – Improved Performance and Expanded Model Support

The second key update focuses on improving the performance and usability of AI-powered auto-labeling. Neurocle previously introduced its transformer-based AI labeling feature, the Auto-Selector, which enables automated labeling through simple clicks and drags. In version 4.4, the underlying AI model has been significantly upgraded to allow for faster and more precise labeling.

Notably, the scope of support has been expanded beyond existing segmentation models to include object detection models. Considering the diverse requirements across manufacturing inspection projects, this enhancement enables a wider range of applications to benefit from AI-driven labeling capabilities.

In addition, a new object size-based threshold setting feature has been introduced to allow users to fine-tune labeling results according to desired object sizes and dimensions. This helps reduce unnecessary labeling of irrelevant objects or background elements, while improving both efficiency and accuracy by focusing only on target areas. The feature is expected to be particularly useful in complex industrial environments where components of varying shapes and sizes coexist.

Meanwhile, Neurocle plans to actively promote the strengths of its Auto Deep Learning vision inspection technology through major domestic and international exhibitions. Starting with AI EXPO KOREA in Seoul this May, the company will expand its outreach to global manufacturing markets by participating in events such as BUTECH (Busan International Machinery Fair) and COMPUTEX in Taiwan. Through these efforts, Neurocle aims to accelerate the adoption of deep learning-based vision inspection technologies in industrial settings and drive their expansion across a wide range of industries.