AI Deep Learning Vision Software Company Neurocle introduces Neuro-T 3.0 & Neuro-R 3.0

With Maximized Labeling Convenience and Multi-model Design

Deep learning vision software company Neurocle has released Neuro-T 3.0 and Neuro-R 3.0, deep learning vision software that can be used without artificial intelligence expertise. Neuro-T and Neuro-R have already been recognized for their excellence after extensive usage in various industries for deep learning vision projects such as defect detection, medical image analysis, and logistics packaging inspection. With Neuro-T and R, users can experience the advantage of automatically generating high-performance models without coding using the Auto Deep Learning Algorithm, then conducting inference images and videos at high-speed with the generated models.

After version 2.3's release in July 2021, which enabled rapid maintenance of deep learning models, Neuro-T 3.0 and Neuro-R 3.0 were officially released on May 30. Neuro-T 3.0 and Neuro-R 3.0 were developed with a focus on model design and labeling convenience. In addition to the Fast Retraining function which is useful when models need to be changed, user convenience has been greatly enhanced.

Data preprocessing, data analysis, model selection, model evaluation, and inference—the entire process of deep learning projects can be done with Neuro-T and Neuro-R. Even partial data processing tasks, which were previously processed using separate libraries such as image cropping and image resizing, is all possible in Neuro-T 3.0 so that users can handle the entire process without leaving Neurocle software.

The core features of the new version of 3.0 are the following.

Auto-Labeling is a feature that provides automatic recommendation for labels based on existing labels. Image labeling is usually the most labor-intensive and time-consuming process of a deep learning project. Through the Auto-Labeling function, the effort and time spent on this phase are greatly reduced, allowing users to go on with the project and create deep learning models more quickly. Along with reduced resources, high-accuracy labeling based on the previous models also guarantees high performance. Even when labeling by hand, the new Magic Wand feature makes the process more efficient by automatically selecting regions of similar colors based on the color value of pixels clicked by the user.

Users can visually structure the entire process before beginning the project, choosing which model to conduct tests with and estimating how many steps it would require to produce the desired model. By implementing functions such as 'labels to images', which extracts labels as images within the Flowchart tab, users do not have to spend as much time on image processing. In addition, users can expect collaboration to advance deep learning projects through the sharing of Flowcharts among team members.

On top of this, movement between tabs, data management, and model training is all faster than the previous version. These improvements, along with even better high-performance models, will increase the efficiency of deep learning projects and user satisfaction.

Hence, Neurocle software is in line with the direction in which the current AI ecosystem should develop. At the 2021 AIGS Symposium held by the Ministry of Science and ICT in December, academic officials such as Seoul National University Professor Seung-Woo Seo and GIST AI Graduate School President Jong-Won Kim stressed that "AI researchers need additional practical software experience to enter the industrial field." Since many industry domains that apply AI technology have experts with decades worth of domain knowledge and experience, they proposed that the expertise of the domain professionals and AI technology should be combined.

From this point of view, Neurocle software provides the best utility for experts in each domain. The Auto Deep Learning Algorithm paired with intuitive graphical user interface (GUI) eliminates the need for users to have deep learning expertise. Since Neurocle software replaces deep learning engineers and vision experts, experts in each domain, such as medical, manufacturing, logistics, and security, can use Neurocle software to carry out image-based projects without restriction.

With version 3.0, Neurocle aims to go beyond improving the software's role as a deep learning expert. Furthermore, Neurocle expects to accelerate the attraction of new customers and expand its business expansion by facilitating project design to establish systematicity and strengthening functions.

Neurocle CEO Hongsuk Lee said, "In version 3.1, we will reinforce the Flowchart to enable image inference from generated models and upgrade functions to make labeling and data management even more convenient."

Neurocle is expanding its business by signing contracts with medical institutions such as Korea University Hospital and the National Institute of Scientific Investigation and various partners such as Renault Samsung, Hyundai, and LG. It is worth staying tuned to the future of Neurocle in the AI deep learning market.