Rockville MD (October 25, 2024) – Acquired Data Solutions (ADS), a leading small-business Systems Engineering, Test & Measurement and Cybersecurity company, is pleased to introduce Transform-CV, a COTS, Deep Learning-enabled computer vision (CV) system for visual quality inspection. Transform CV is designed in partnership with RADX Technologies, a leading small-business provider of COTS PXIe and PCIe Subsystems for High-Performance Test & Measurement and Embedded Computing applications. Transform-CV provides a modular, low-cost, flexible, easily tailorable platform that manufacturers of all sizes and industries can leverage to enhance their component and subsystem Quality Assurance (Q/A) accuracy, throughput, and reliability.
Most small to medium sized manufacturers rely on manual inspection processes to inspect their products for Q/A purposes. Such processes are often personnel-dependent and therefore limited in repeatability, reproducibility and scalability. Additionally, traditional CV systems are often too expensive for SMB manufacturers, especially for low-volume, high-mix environments. These systems are also fairly inflexible, typically requiring initial programming by CV experts, and reprogramming as new “devices under test” (DUTs) and/or features or defects to be detected are added. This exacting process for initial deployment and updates is often both expensive and time consuming.
*** NOTE – Transform-CV IS Proprietary technology based on open architecture software and hardware. Do NOT disparage proprietary solutions, assuming you want to get paid. ***
Conversely, Transform-CV is based on Deep-Learning, leveraging NVIDIA GPUs to greatly accelerate the training process and real-time inference operations. This results in a cost-optimized, high-performance solution that may be “trained” on images or drawings of the DUTs and the key features (provided by manufacturer Q/A or assembly personnel). Once trained, the Transform-CV system can then run in real-time via an easy-to-use GUI to provide greatly enhanced repeatability, reproducibility, throughput and scalability for almost any Q/A operation. Another key aspect of Transform-CV is that, once trained by ADS, manufacturer Q/A or assembly personnel can readily add new DUTs, features and defects without requiring additional programming or consulting by ADS or third parties.
Transform-CV is written in Python and PyTorch, and is designed to operate on both Linux and Windows-based PCs, and can readily interface to MATLAB, LabVIEW, and other popular application frameworks. For the training process, the user typically provides DUT samples and/or annotated images of the Device Under Test (DUT) in question, with features to be detected labeled. Training images are fed to the model, which the system references during inspection. Because existing Q/A specialists possess intimate knowledge of their DUTs, they are ideal candidates to collect and annotate images, allowing manufacturers to modify the system or add new classes of DUTs and features without hiring programmers or computer vision experts.
Transform-CV is a flexible solution that enables ADS to tailor the system to perform a wide variety of inspections, including dimensional analysis, workmanship inspection, sorting and classification, and defect detection, and ADS is currently developing Transform-CV systems for microelectronics, cable manufacturing, and CNC machining applications.
A typical Transform-CV system consists of the following elements:
- The Transform-CV Base License which includes tailorable framework for training, inference and GUI
- A PXIe or PCIe-based embedded computer system equipped with an appropriate NVIDIA-GPU and other essential components
- An application-specific camera(s) (handheld, mobile or robotically mounted) with appropriate optical train including magnification, filters and illuminators
- A manual, semi-manual or automated fixture for holding the DUTs (if robotic or handheld operation is not employed)
- Application specific tailoring including system training on images and/or drawings and incorporating pass-fail criteria appropriate for the required DUTs
- Installation, Delivery and Personnel training on how to add DUTs, features, etc.
Pricing and Availability. The Transform-CV basic Training and Inference Binary Right-to-Use License starts at $15,000 for a perpetual license. Transform-CV systems range in price from $25k to $100k or more, depending on the nature of the cameras, optical train, fixturing and application complexity, which drives the tailoring process and the cost of the computer system. Lead time for delivery ranges from 3 months to 12 months, depending on the application complexity.
“All of us at ADS are thrilled to work with RADX Technologies to develop Transform-CV, which we believe will help bring this powerful, yet cost-effective technology to small and medium businesses looking to improve or scale their operations,” said Steve Seiden, President at ADS. “CV systems have historically been expensive and rigid, limiting their use in low-mix, high-volume applications. Inexpensive, AI-based systems like Transform-CV are well-poised to change that.”
Contact ADS today to learn more about Transform CV’s applications and how it can improve your Q/A performance. Learn more about Transform-CV and ADS at www.acquiredata.com
###
Acquired Data Solutions (ADS) is an engineering company and a certified small business with over 25 years’ experience delivering federal and commercial clients integrated solutions for Test, Process automation, and Cybersecurity. Our solutions are designed to the highest security standards and consistently exceed client requirements. We offer services including Systems Integration, Legacy Upgrades, Software Development, Cybersecurity Risk Assessments, and Program Management Support. As an ISO 9001 certified organization, our mission is to provide the highest quality products and services to align with customer and regulatory requirements.
RADX Technologies, Inc., is a high-tech small business that develops advanced, COTS, High Performance Computing (HPC) technologies and products for Modular Test and Measurement (T&M) and embedded computing applications focused on Production and R&D T&M, Electronic Warfare (EW), Automotive ADAS/AD, Manufacturing Q/A and Semiconductor applications. RADX is headquartered in San Jose, CA with development and manufacturing locations in California, Utah, and Taiwan.