Neurocle, a company specializing in AI deep learning vision inspection software, announced that it will participate in the AI EXPO KOREA 2025 to be held from May 14 to 16 at COEX in Seoul, where it will showcase its auto deep learning vision inspection technology.

AI EXPO KOREA is the largest artificial intelligence exhibition in Korea, serving as a platform for industry professionals across various sectors considering the adoption of AI technologies to explore the latest solutions and trends. At this year’s exhibition, Neurocle will present a range of vision inspection technology demonstrations tailored to the manufacturing industry, centered on its deep learning model training software Neuro-T and its runtime library Neuro-R, designed for seamless integration with vision inspection systems. It is notable that with the company’s auto deep learning algorithm which automatically optimizes model hyperparameters to generate high-performance models, Neurocle enables even non-experts to easily build AI-powered inspection models.

Neuro-T provides one of the industry’s most extensive selections of deep learning models, allowing it to adapt to a wide range of inspection environments. It also supports AI-based automatic labeling functionalities, significantly reducing the time and cost required compared to manual labeling.

The platform offers a comprehensive range of models capable of flexibly addressing diverse inspection needs within a single environment. These include fundamental models such as image classification, defect region segmentation, and object detection, as well as more advanced capabilities such as synthetic defect generation with GAN Defect Generator technology, anomaly detection based on unsupervised learning, and patch classification for detecting subtle defects in high-resolution images.

Its automated labeling features also stand out. Foundation model-based Auto-Selector recommends regions with similar characteristics after the user labels only a small number of images. The Keyword Labeler identifies and marks relevant objects within images using bounding boxes based on short keyword inputs, while the Shape Converter automatically transforms these box-type labels into more precise polygons. Together, these functions minimize the time and labor required for labeling while acquiring refined and consistent training data.

A Neurocle official stated, “Through our auto deep learning technology, anyone can implement AI-based vision inspection quickly and accurately,” adding, “At this exhibition, we will demonstrate the value of practical AI inspection solutions that can be immediately deployed on manufacturing sites.”

Meanwhile, Neurocle recently released version 4.4 of its software. The update introduces support for the NVIDIA RTX 50 series, enabling optimized operation of deep learning models on the latest high-performance hardware, as well as faster and more precise AI labeling capabilities. In addition, a new object size-based threshold setting feature has been added, allowing stable performance in complex inspection environments.

Following its participation in AI EXPO KOREA 2025, Neurocle plans to accelerate its global market expansion by consecutively taking part in major domestic and international exhibitions, including BUTECH 2025 (The 12th Busan International Machinery Fair) in Busan and COMPUTEX 2025 in Taiwan.