Neuro-T & Neuro-R 3.2 Released!

Neurocle has recently released Neuro-T & Neuro-R 3.2. Neurocle's AI deep learning solution is drawing attention from the manufacturing industry as it is presented as a groundbreaking alternative to rule-based machine vision inspection or visual inspection by humans. Generally, it is regarded that the rule-based machine vision requires frequent and complex rule programming. For the visual inspection, the biggest challenge is frequent false and over-examination due to variation in inspector’s condition.

Neuro-T is a deep learning model trainer where users can implement image data preprocessing, build AI deep learning model, and evaluate the performance of models. Neuro-T offers six types of deep learning models including ‘Image Classification’, ‘Object Detection’, and ‘Segmentation’. Among these options, users can choose a single model that is most suitable for the project or combine multiple models for a complex project.

Neuro-R is a runtime API where users can deploy the model they have created with Neuro-T to manufacturing sites. Through Neuro-R, image or video inference is implemented in real-time.

▲Neuro-T & Neuro-R v3.2

Major updates are as follows.

Users can now interchange the compatible labeling files between ‘Object Detection Model’ and ‘Segmentation Model’. It is designed to dramatically reduce labeling resources by making the most of existing labeling when changing model types. For example, if labeling is completed with an Object Detection model, the labeling file of the model can be extracted and inserted in the form of box labeling into a Segmentation model based on the same image set. It is same for the vice versa. This feature is useful for users who wish to experiment with various model types to generate deep learning models with higher performance.

Neuro-R v3.2 supports the Nvidia GPU 40 series. Users can use the latest GPU to significantly reduce their inference time. In addition, the option to set the input size of the image up to 1024x1024 was added. Thanks to the ability to insert (import) high-resolution images, users can create high-performance models and detect fine defects with high accuracy.

A grouping function has also been added to enable more convenient and systematic management of image datasets. Through this function, a large amount of image data can be classified into various groups and groups can be sorted through filtering.

A system that can efficiently manage multiple projects has also been added. Using this system, users may flexibly change the priority of the learning progress for each model according to the timeliness of the project. "In the manufacturing industry, it is the competitiveness of the industry to quickly identify the cause of defects and increase yields through an advanced inspection system. As a result, inquiries regarding the introduction of AI deep learning vision tests are rapidly increasing.” said Hongsuk Lee, CEO of Neurocle. "We will continue to upgrade our products so that more customers can acquire quality competitiveness through Neurocle's solutions." he added.

Neurocle specializes in AI deep learning vision software for vision inspection in manufacturing industries. Neurocle’s software guarantees the industry's highest level of inspection accuracy, loved by global mega companies like LG, SK, and Hyundai. Neurocle is currently actively targeting global markets by conducting business with a total of 15 Asian and European countries.