The SMore ViMo series products are equipped with intelligent production models for over 1000 segmented industry scenarios, embedded with over 50 tools such as algorithm enhancement, and can provide full stack intelligent capabilities for the entire manufacturing process of the industry. With the help of the SMore ViMo series of industrial products, a cross industry intelligent manufacturing ecosystem can be quickly built. SMore ViMo covers industrial intelligent clouds, deep learning training software, machine vision software, and other industrial products, providing data management, visual solution design, end-to-end running software, and other software to meet the visual needs of industry. The factory can quickly implement the entire process of model training, software integration, and production line deployment without the involvement of professional algorithm personnel, greatly reducing the threshold for the use of intelligent applications in the manufacturing industry and providing strong inclusive and intelligent capabilities for industry.
SMore ViMo provides data management, visual solution design, end-to-end running software, and more to meet the visual needs of the industry. The factory can quickly implement the entire process of model training, software integration, and production line deployment without the professional algorithm personnel, greatly reducing the threshold for the use of intelligent applications in the manufacturing industry and providing intelligent capabilities for industry. The Smore ViMo product matrix currently includes ViMo Cloud, ViMo Deeplearning, and ViMo Studio.
The project requires a 3-class classification (OK/NG/NG-2) of scratches and dirt on the USB interface. Previously, the client's quality inspection was mainly conducted through manual visual inspection, which resulted in high costs and low detection efficiency. Additionally, the tested products have various defect types, requiring precise defect classification and statistics based on their characteristics. Traditional vision algorithms cannot define standard detection templates for certain defects, necessitating the use of high-precision deep learning algorithms for detection. Based on SMore ViMo's high-precision algorithm capability, it enables various defect detection with both the accuracy rate and recall rate reaching 98.9%, and the false negative rate below 2%, which empower full automation in the production line, thus leading to low cost, improved product yield and efficient operation.
The battery quality inspection process of a leading new energy company mainly adopts a combination of automated glue coating and manual visual inspection. The drawback of this approach is its high cost and low production efficiency. Therefore, the customer hopes to further enhance the quality of inspection in terms of unmanned operation, high efficiency, and high accuracy. We adopt the solution design system of Vi-Studio from SMore ViMo, and utilize a high-precision algorithm to achieve various defect detections, such as leakage glue detection, scatter point detection, and broken glue detection. Vi-Lab operation management system is able to achieve machine vision delivery both swiftly and directly, which makes it convenient for customers to deploy the production line quickly. This also speeds up production space, enabling the cycle time reduction of the photo-calculation-material dispensing action to within 500ms per piece, which can satisfy customers' requirements of time, then boost the automation of production line and smart manufacturing.
Previously, this enterprise only adopt a manual sampling inspection method to defect silicon wafer character, which proved to be low-efficient and the results are driven by subjective. During the inspection, the on-site lighting may be dim, resulting in poor image quality. Additionally, the characters being inspected may have certain angles of curvature, requiring precise positioning. SMore ViMo smart industrial platform is equipped with optical character recognition, defect segmentation, and defect detection modules. It has a 100-million-pixel resolution algorithm capable of processing high-resolution images. It features intelligent semantic matching, which can solve complex search problems in large-scale chips. With dynamic data augmentation and few-shot learning, it accelerates the learning process and handles complex detection tasks, enabling the enterprise to achieve intelligent full inspection of silicon wafers.
This project involves recognizing characters engraved on metal components of smartwatches. There are three different fonts for the characters, and the watchbands come in three colors: black, white and orange. The imaging conditions are complex, including blurriness, glare, and tilting, among other variations, resulting in a total of 15 different forms. This presents significant challenges for character recognition.
Based on SMore ViMo industrial OCR algorithm, SmartMore has implemented a complex and efficient pre-processing method for characters. This enhances the algorithm's robustness to different environmental conditions, resulting in a character recognition accuracy up to 99.9%. The solution is compatible with various product forms and has been successfully deployed in the fully automated production of six production lines.
The characters engraved on the metal components of vibrating motors in mobile phones often appear blurry and distorted, and the reflective nature of the metal further complicates the imaging process, making them difficult to recognize. These components are often smaller than the size of a fingernail, and the characters engraved on them are as small as 0.1, making them challenging for human eyes to identify. Additionally, the characters can be oriented in various directions and have different engraving styles, further adding to the difficulty of recognition.
The client wishes to introduce machine vision technology for visual defect detection in the production process, aiming to improve product quality and reduce repair rates. The project involves detecting over 20 different types of defects, covering various areas and categories. High accuracy is required for the detection process. Based on the SMore ViMo Intelligent Industrial Platform, SMore has developed a multi-classification detection algorithm that utilizes deep learning techniques to optimize the overall performance. This algorithm achieves pixel-level processing, ensuring precision down to each individual pixel. With a processing speed of 0.2 seconds per piece, it enables 100% coverage of the production line with only 20% of the quality inspection workforce.
The manufacturer's bearings have extremely small defects, such as scratches, that are not visible to the naked eye, requiring high accuracy from the algorithm. There are over 20 different types of defects, and their distribution locations are relatively random, making it challenging to locate and identify them.
To address the customer's pain points, SMore adopts the SMore ViMo intelligent industrial platform, along with an optical solution equipped with four types of light sources, hardware configuration of edge computing integrated machine, color camera, and 4K color line-scan camera, to achieve an integrated solution for automotive bearing inspection. The SMore ViMo platform, equipped with high-precision algorithms, can recognize over 23 different types of defects in different product models, such as rust, oil stains, scratches, and so on, thus leading to improved precision and high quality inspection detection rate.
The customer's existing quality inspection process for lithium batteries primarily relies on manual self-inspection, resulting in high costs and low production efficiency, with a persistently high rate of missed defects. Therefore, the customer wishes to further enhance quality inspection in terms of automation, efficiency, and accuracy to improve production efficiency, product qualification rate, and environmental standards.
SMore ViMo and Vi-Lab systems are utilized by SmartMore to develop customized visual inspection software for defect detection in the product. The software enables multiple defect detections, achieving a product detection rate of ≥99.9%. The hardware solution is based on a commonly used multi-camera sorting machine in the inspection industry, serving as a labeling machine. The processing speed can reach up to 7200pcs/h. In addition, the platform seamlessly integrates with integrated equipment, and the Vi-Lab operation management system facilitates quick delivery of machine vision projects, allowing customers to deploy production lines rapidly.
The middle frame of computer requires the detection of 15 types of defects, including nut installation, water ports, camera foam adhesive, etc. Previously, due to the metal material of the middle frame being prone to reflections and poor image quality, combined with inconsistent detect definitions in the early stage, manual defect detection was time-consuming and had low accuracy, impacting product production efficiency.
Using the Vi-Studio solution design system in the SMore ViMo intelligent industrial platform, SmartMore has refined the classification, detection, and segmentation algorithms into an integrated algorithm model. These algorithms are specifically designed to detect different types of defects, addressing the customer's specific inspection requirements. The algorithms have been enhanced to improve robustness in various environmental conditions, achieving a defect detection rate of ≥99.9%. Additionally, the Vi-Lab operation management system in SMore ViMo enables swift delivery of machine vision projects. From understanding the requirements to deploying the final product, the entire project can be completed within 28 days.
In the production process of diodes, there are many defects such as fractures that affect the performance of the diodes. Detecting these surface defects has always been a major challenge for the manufacturing company. Previously, the client used a traditional computer vision algorithm solution, which had a high rate of false negatives and false positives, requiring additional manpower for manual verification, significantly increasing the company's costs. Based on the segmentation algorithm of SMore ViMo, SmartMore can perform the detection of over 10 types of defects on the product, including fractures, bubbles, shallow edge cracks, and more. The segmentation accuracy of the algorithm has surpassed that of traditional algorithm detection and manual inspection. It can also provide information on the area and location of the defects, achieving a product detection rate of 81% or higher, a false positive rate of 2% or lower, and a false negative rate of 0.01% or lower. This solution helps the customer save over 3 million yuan in costs annually.
