Neurocle: "User-Friendly Deep Learning Vision Inspection Solutions, Neuro-T and Neuro-R"
Neurocle: "User-Friendly Deep Learning Vision Inspection Solutions, Neuro-T and Neuro-R"
Posted December. 23, 2024 15:19,
Updated December. 26, 2024 17:42
Neurocle develops deep learning-based vision inspection solutions, Neuro-T and Neuro-R, that are easy for even non-experts to use. Neuro-T allows users to effortlessly train and create vision inspection models, while Neuro-R enables the real-time operation of the models created with Neuro-T.
Neurocle’s solutions support nine types of inspection models and are widely applied across a range of industries that includes secondary batteries, semiconductors, displays, automobiles, food, and university hospitals. The company has also entered overseas markets, providing its solutions in more than 25 countries, including in Asia, Europe, and the United States. Neurocle attributes its success to its proactive approach of listening to customer feedback and addressing their needs.
We spoke with Hong-suk Lee, CEO of Neurocle, to discuss the company, its Neuro-T and Neuro-R solutions, and its commitment to making AI technology more accessible through ongoing research and global market expansion.
Hong-suk Lee, CEO of Neurocle / Source: IT DongA
Developing a Standard Toolkit Software: The Founding of Neurocle
IT DongA: Hello, CEO Lee. Can you please introduce yourself for our readers?
Lee: Hello, my name is Hong-suk Lee and I am the CEO of Neurocle. I have been involved in AI and deep learning since 2016, when these technologies began gaining attention. My work has been focused on solving technical challenges in different fields of industry, and expanding the practical applications of AI. I’ve worked on projects with large manufacturing companies, achieved meaningful results in global markets, and contributed to building innovative business models through strategic planning and market development. Currently, I am dedicated to developing deep learning-based vision inspection solutions with the Neurocle team that will enhance efficiency across multiple industries.
IT DongA: What inspired you to establish Neurocle?
Lee: I used to work for a company that developed AI and deep learning solutions for clients. But I often felt that by creating new solutions for every client, we were using too many resources. I realized that providing a standardized toolkit software instead of custom solutions for each client could make the process more efficient for suppliers. If the software performed well, clients would also be satisfied. This idea led me to establish Neurocle in 2019.
IT DongA: Could you introduce Neurocle for our readers?
Lee: Neurocle operates under the vision of "Making Deep Learning Vision Technology More Accessible." We strive to create an environment in which anyone can easily use AI. To achieve this, we developed deep learning-based vision inspection solutions, Neuro-T and Neuro-R.
The company name, Neurocle, combines the words "neural" from "neural network," the core of deep learning technology, and "miracle," to represent the creation of new possibilities through neural networks.
Deep Learning Vision Inspection Solutions: Neuro-T and Neuro-R
IT DongA: Please tell us about Neuro-T and Neuro-R.
Lee: Neuro-T trains AI models, and Neuro-R operates the models trained with Neuro-T in real time. "T" stands for "Training," and "R" stands for "Runtime."
Neurocle’s deep learning vision inspection solutions / Source: Neurocle
Neuro-T is a solution that allows users to easily train and create deep learning vision inspection models. Users upload image data to Neuro-T, label the data using various tools, and train and create models based on the labeled data.
Neuro-T employs an Auto Deep Learning Algorithm that automatically optimizes models to suit the client’s specific field. This eliminates the need for deep learning engineers to manually optimize model structures and hyperparameters, which allows users to avoid additional customization or optimization steps.
The ease of use is another hallmark of Neuro-T. Its no-code, GUI-based interface means that users with no technical knowledge can create high-performance models with just a few clicks. This simplicity has made it easy for practitioners in medical institutions and manufacturing sites to use.
Neuro-R applies the models created with Neuro-T in real time at manufacturing and medical sites. Its modular library system enables it to seamlessly integrate with various equipment and systems. For example, in manufacturing, Neuro-R can be integrated with inspection devices, while in healthcare, it can be linked to medical software. This flexibility allows customers to use the solution freely, without being restricted by their existing systems or project environments.
IT DongA: What industries do Neurocle’s solutions target?
Lee: Vision inspection has significant potential in the manufacturing and medical industries. Since our solution is a standard toolkit software, its target industries are broad. In the manufacturing industry, it is used for visual inspections to detect product defects by analyzing images captured by cameras. In the medical field, it supports research using images captured by equipment such as X-rays, MRIs, CT scans, endoscopes, and ultrasounds.
Neurocle solutions are widely used in the manufacturing and medical fields / Source: Neurocle
Currently, we support nine models: ▲ ‘Classification,’ which classifies images; ▲ ‘Patch Classification,’ which segments patches to determine normalcy; ▲ ‘Segmentation,’ which detects defect shapes and locations at the pixel level; ▲ ‘Object Detection,’ which identifies objects and their positions; ▲ ‘Defect Generator,’ which creates virtual defect images; ▲ ‘OCR,’ which recognizes text in images; ▲ ‘Rotation,’ which rotates images to the correct orientation; and ▲ ‘Anomaly Classification/Segmentation,’ which learns normal images in order to detect abnormal ones.
Recently added models include Patch Classification and Defect Generator. These were developed to address common issues in manufacturing, such as the need for high-resolution image processing and a lack of defect image data. Patch Classification divides high-resolution images into patches, ensuring high detection accuracy. Defect Generator creates realistic defect images, allowing model training even when defect image data is scarce.
Using Neurocle’s solutions, manufacturing industries can automate defect inspections that were previously reliant on human labor, increasing inspection speed and reducing costs while minimizing human error. In the medical field, the solutions can enhance the speed and efficiency of research.
Continuous R&D Domestically, Strengthened Presence Overseas
IT DongA: Could you tell us about the progress of your business so far?
Lee: We released the first version of our solution in 2020. Since then, we have added new models and enhanced their features, launching new versions more than three times annually.
Our solutions are currently being provided to leading manufacturing and medical companies, both domestically and internationally. We have clients in diverse sectors, including secondary batteries, semiconductors, displays, automobiles, tires, food, university hospitals, and general hospitals.
Right after launching the product, we also pursued global market entry. Since our solution is a standard toolkit software that does not require customer-specific customization, it is highly exportable. From the beginning, we focused on expanding globally, and we now provide our solutions to more than 25 countries across Asia, Europe, and the United States.
We listen carefully to our customers, gather their feedback, prioritize it, and implement it. As well, we proactively propose solutions to the issues they raise. As a result, our solutions have become products that meet customer needs and bring them a high level of satisfaction, which I believe is one of the factors behind our success.
Hong-suk Lee explaining Neuro-T and Neuro-R / Source: IT DongA
IT DongA: How has SBA supported your business?
Lee: Neurocle has participated in various SBA support programs, which provided practical assistance in business operations and technology development. One of the most helpful initiatives was the open innovation program with large companies. SBA facilitated connections between technology providers like us and the companies that needed these solutions. They also supported proof-of-concept (PoC) projects with funding and resources. Through these projects, we collaborated with partner companies, directly identified their needs, and strengthened the competitiveness of our technology and solutions.
IT DongA: What are Neurocle’s future plans and goals?
Lee: On the R&D front, we will continue to enhance the functionality of Neuro-T and Neuro-R to improve their usability. Market trends and customer needs are constantly evolving, so we are preparing new features and models to meet these demands. Version 4.3 is scheduled for release soon, and we plan to roll out further updates in line with our roadmap.
In terms of global expansion, we aim to strengthen our presence in international markets. While we already have a significant share in global markets, there are still untapped countries and industries with potential. We are exploring various strategies to enter these markets.
Our ultimate goal is to democratize AI technology. Through ongoing research and global market expansion, we aim to deliver our innovative technology and solutions to diverse countries and industries.
by Man-hyuk Han (mh@itdonga.com)
Neurocle’s solutions support nine types of inspection models and are widely applied across a range of industries that includes secondary batteries, semiconductors, displays, automobiles, food, and university hospitals. The company has also entered overseas markets, providing its solutions in more than 25 countries, including in Asia, Europe, and the United States. Neurocle attributes its success to its proactive approach of listening to customer feedback and addressing their needs.
We spoke with Hong-suk Lee, CEO of Neurocle, to discuss the company, its Neuro-T and Neuro-R solutions, and its commitment to making AI technology more accessible through ongoing research and global market expansion.
Developing a Standard Toolkit Software: The Founding of Neurocle
IT DongA: Hello, CEO Lee. Can you please introduce yourself for our readers?
Lee: Hello, my name is Hong-suk Lee and I am the CEO of Neurocle. I have been involved in AI and deep learning since 2016, when these technologies began gaining attention. My work has been focused on solving technical challenges in different fields of industry, and expanding the practical applications of AI. I’ve worked on projects with large manufacturing companies, achieved meaningful results in global markets, and contributed to building innovative business models through strategic planning and market development. Currently, I am dedicated to developing deep learning-based vision inspection solutions with the Neurocle team that will enhance efficiency across multiple industries.
IT DongA: What inspired you to establish Neurocle?
Lee: I used to work for a company that developed AI and deep learning solutions for clients. But I often felt that by creating new solutions for every client, we were using too many resources. I realized that providing a standardized toolkit software instead of custom solutions for each client could make the process more efficient for suppliers. If the software performed well, clients would also be satisfied. This idea led me to establish Neurocle in 2019.
IT DongA: Could you introduce Neurocle for our readers?
Lee: Neurocle operates under the vision of "Making Deep Learning Vision Technology More Accessible." We strive to create an environment in which anyone can easily use AI. To achieve this, we developed deep learning-based vision inspection solutions, Neuro-T and Neuro-R.
The company name, Neurocle, combines the words "neural" from "neural network," the core of deep learning technology, and "miracle," to represent the creation of new possibilities through neural networks.
Deep Learning Vision Inspection Solutions: Neuro-T and Neuro-R
IT DongA: Please tell us about Neuro-T and Neuro-R.
Lee: Neuro-T trains AI models, and Neuro-R operates the models trained with Neuro-T in real time. "T" stands for "Training," and "R" stands for "Runtime."
Neuro-T is a solution that allows users to easily train and create deep learning vision inspection models. Users upload image data to Neuro-T, label the data using various tools, and train and create models based on the labeled data.
Neuro-T employs an Auto Deep Learning Algorithm that automatically optimizes models to suit the client’s specific field. This eliminates the need for deep learning engineers to manually optimize model structures and hyperparameters, which allows users to avoid additional customization or optimization steps.
The ease of use is another hallmark of Neuro-T. Its no-code, GUI-based interface means that users with no technical knowledge can create high-performance models with just a few clicks. This simplicity has made it easy for practitioners in medical institutions and manufacturing sites to use.
Neuro-R applies the models created with Neuro-T in real time at manufacturing and medical sites. Its modular library system enables it to seamlessly integrate with various equipment and systems. For example, in manufacturing, Neuro-R can be integrated with inspection devices, while in healthcare, it can be linked to medical software. This flexibility allows customers to use the solution freely, without being restricted by their existing systems or project environments.
IT DongA: What industries do Neurocle’s solutions target?
Lee: Vision inspection has significant potential in the manufacturing and medical industries. Since our solution is a standard toolkit software, its target industries are broad. In the manufacturing industry, it is used for visual inspections to detect product defects by analyzing images captured by cameras. In the medical field, it supports research using images captured by equipment such as X-rays, MRIs, CT scans, endoscopes, and ultrasounds.
Currently, we support nine models: ▲ ‘Classification,’ which classifies images; ▲ ‘Patch Classification,’ which segments patches to determine normalcy; ▲ ‘Segmentation,’ which detects defect shapes and locations at the pixel level; ▲ ‘Object Detection,’ which identifies objects and their positions; ▲ ‘Defect Generator,’ which creates virtual defect images; ▲ ‘OCR,’ which recognizes text in images; ▲ ‘Rotation,’ which rotates images to the correct orientation; and ▲ ‘Anomaly Classification/Segmentation,’ which learns normal images in order to detect abnormal ones.
Recently added models include Patch Classification and Defect Generator. These were developed to address common issues in manufacturing, such as the need for high-resolution image processing and a lack of defect image data. Patch Classification divides high-resolution images into patches, ensuring high detection accuracy. Defect Generator creates realistic defect images, allowing model training even when defect image data is scarce.
Using Neurocle’s solutions, manufacturing industries can automate defect inspections that were previously reliant on human labor, increasing inspection speed and reducing costs while minimizing human error. In the medical field, the solutions can enhance the speed and efficiency of research.
Continuous R&D Domestically, Strengthened Presence Overseas
IT DongA: Could you tell us about the progress of your business so far?
Lee: We released the first version of our solution in 2020. Since then, we have added new models and enhanced their features, launching new versions more than three times annually.
Our solutions are currently being provided to leading manufacturing and medical companies, both domestically and internationally. We have clients in diverse sectors, including secondary batteries, semiconductors, displays, automobiles, tires, food, university hospitals, and general hospitals.
Right after launching the product, we also pursued global market entry. Since our solution is a standard toolkit software that does not require customer-specific customization, it is highly exportable. From the beginning, we focused on expanding globally, and we now provide our solutions to more than 25 countries across Asia, Europe, and the United States.
We listen carefully to our customers, gather their feedback, prioritize it, and implement it. As well, we proactively propose solutions to the issues they raise. As a result, our solutions have become products that meet customer needs and bring them a high level of satisfaction, which I believe is one of the factors behind our success.
IT DongA: How has SBA supported your business?
Lee: Neurocle has participated in various SBA support programs, which provided practical assistance in business operations and technology development. One of the most helpful initiatives was the open innovation program with large companies. SBA facilitated connections between technology providers like us and the companies that needed these solutions. They also supported proof-of-concept (PoC) projects with funding and resources. Through these projects, we collaborated with partner companies, directly identified their needs, and strengthened the competitiveness of our technology and solutions.
IT DongA: What are Neurocle’s future plans and goals?
Lee: On the R&D front, we will continue to enhance the functionality of Neuro-T and Neuro-R to improve their usability. Market trends and customer needs are constantly evolving, so we are preparing new features and models to meet these demands. Version 4.3 is scheduled for release soon, and we plan to roll out further updates in line with our roadmap.
In terms of global expansion, we aim to strengthen our presence in international markets. While we already have a significant share in global markets, there are still untapped countries and industries with potential. We are exploring various strategies to enter these markets.
Our ultimate goal is to democratize AI technology. Through ongoing research and global market expansion, we aim to deliver our innovative technology and solutions to diverse countries and industries.
by Man-hyuk Han (mh@itdonga.com)
About Dong-A Ilbo