-
Deep Learning-Based
Vision Inspection Platform
-
Vision InspectionAutomate your vision inspection tasks with the powerful, reliable, and easy-to-use SaigeVision® platform. Employing the state-of-the-art in machine learning algorithms, with advanced tools for data generation, augmentation, and labelling, SaigeVision® offers unmatched accuracy and speed for the most challenging inspection problems. Even with only limited training data, SaigeVision® achieves maximum performance with minimum operational overhead.
-
Video Anomaly DetectionSaigeVAD® automatically detects a diverse range of assembly and manufacturing errors from video data. Using only unlabelled training data — a continuously recorded video sequence of normal operations, for example — SaigeVAD® can be trained to detect all anomalies, and to classify which of these are critical and which can be safely ignored.
-
OCR | Text UnderstandingSaigeOCR® automatically detects words and text in images, and transforms these into digital strings. Drawing upon advanced deep learning methods to extract curvature and other geometric information of the text area, SaigeOCR® robustly recognizes and transforms even highly distorted text into digital strings.
-
Signal AnalysisSaige offers solutions for inspection and monitoring of signal and time series data: electrical measurement signals, acoustic signals, vibration patterns, and various multidimensional time series and sequential data.
Industries
-
Energy Storage Systems
Typical battery casing defects such as scratches, dents, cracks, and stains are difficult to detect with conventional machine vision: not only do the reflective surfaces make the defects difficult to capture, but the threshold for defects is also difficult to quantify. Leading EV battery makers like Samsung SDI rely on SaigeVision® to solve their most critical battery inspection needs.
-
Semiconductors and Electronics
Printed circuit boards (PCBs), with their wide variety of defects and large image sizes, pose especially difficult challenges for in-line inspection, with very stringent accuracy and speed requirements. SaigeVision® is currently being used at a major electronics manufacturer (Daeduck Electronics) for fast and accurate in-line inspection of PCBs.
-
Packaging and Logistics
From inspection of defects to identifying and locating features and selecting grasp points for robotic pick-and-place, vision is essential to the packaging and logistics industry. SaigeVision® has been deployed in diverse automated production lines for both in-line inspection and robot grasp point selection.
-
Displays and Panels
OLED/LED/LCD display panels present unique challenges for defect inspection. MURA defects, which can be described as extremely low-contrast stains ranging in size from just a few pixels to large subregions of the panel, are particularly difficult to detect and classify. SaigeVision® has been successfully applied in detecting a wide array of MURA and other defect types in OLED panels.
-
Pharmaceutical and Biomedical
With its stringent safety requirements, inspection of pharmaceutical products cannot tolerate any false negatives (i.e., incorrectly classifying damaged pills as normal), while at the same time it is crucial to maintain real-time, high-volume inspection rates. The advanced algorithms employed by SaigeVision® are the perfect tool for meeting the strict accuracy and speed requirements of the pharmaceutical industry.
Tech Blog
-
생산된 제품의 결함 검사에서 더 나아가, 결함을 발생시키는 공정까지 포착할 수 있다면 어떨까요? 원인 공정을 찾아내어 결함 발생률 자체를 줄일 수는 없을까요? 이런 문제를 해결하기 […]Taeseung Baek -
세이지리서치의 미션: Inspection Done Right 모든 제조업 현장에서는 생산된 제품의 하자 여부를 검사하는 공정을 반드시 거칩니다. 이 공정을 ‘QC (quality control) 공정’이라 칭하는데, 기계가 […]Seunghyeon Kim -
Date: 2021.07.20. Writer: Hyeokjun Kwon Source: Saige Blog – A Peek into Automatic Data Augmentation by Policy Searching Machine-learning(ML, 기계학습) 혹은 Deep-learning(DL, 딥러닝) 기반으로 […]Hyeokjun Kwon -
Date: 2021.07.08 Writer: Seunghyeon Kim Source: Saige Blog – Transfer Learning for Industrial Visual Inspection Transfer learning 이란? 2012년 Alex Krizhevsky가 AlexNet[1]이라는 convolutional neural […]Seunghyeon Kim