Defect Visual Inspection (DVI) is an AI-powered solution for visual anomaly detection at the edge.
It offers a quick setup process and automatically adapts to new tasks and images, making it a versatile option for a variety of use cases. This flexibility makes DVI particularly well-suited for OEMs, system integrators, and large corporate users looking for robust and efficient inspection tools.
Is anomaly detection the right tool for your problem?
Anomaly detection is beneficial if the defect is hard to enumerate or define using strict rules. In such cases, it’s clear that something is “off,” even if it’s challenging to describe precisely
Anomaly detection is less suitable when the defect must be accurately measured, counted, or segmented. In these scenarios, DVI should be integrated with other computer vision techniques. Our consulting services can assist with these more specialized needs.
Here are some examples, borrowed from real applications of DVI on the field, of defects that are tough to describe in terms of fixed rules, yet are immediately recognizable as flawed. This is exactly where AI defect detection and DVI excel.
KEY FEATURES
Different anomaly detection algorithms have unique strengths, and DVI leverages these strengths through:
Simplicity
Requires only a small number of nominal training images (e.g., 20) and minimal parameter adjustments. No images of defects or anomalies are needed.
Speed
Training completes in just a few seconds… and you are ready to go!
Edge deployment
By training our AI anomaly detection at the edge, on the device: the visual gap between training and testing samples is minimized; all data remains onsite and private; no additional IT infrastructure (computing, networking, etc); no ongoing fees for external services.
Multi-model capability
Quick switch between different AI models to inspect various parts of the same object or the same part at different scales.
Continuous improvement
With operator feedback, the model can refine its performance over time.
Combining all these features into a single deep learning anomaly detection model is what gives DVI a distinct technical edge over other solutions. It empowers users to drastically cut both time-to-market and the level of post-installation support required, helping you deliver solutions to your clients more efficiently and reliably.
EXAMPLES OF DVI IN ACTION
Our AI visual inspection solution has been successfully deployed across a range of industries, from manufacturing or packaging to food, demonstrating its ability to inspect both fine details and broader surfaces simply by adjusting the input image and resolution.
Whether the case, the configuration remains straightforward. Below are some real-world applications showcasing the effectiveness of DVI.
Detected anomalies on these images of pizza:
- Burnt areas
- Non-uniform distribution of ingredients
- Unexpected ingredients
Detected anomalies on these images of carbon fiber textile:
- Fibers exiting the texture
- Uneven distribution of glue
- Wrong patterns
Detected anomalies on these images of boxes with a transparent plastic wrap:
- Wrinkles on plastic sheet
- Damaged box corners
- Holes on plastic sheet
HOW TO GET IT
There are several ways to integrate our AI quality inspection into your product, depending on your specific requirements.
We offer a range of hardware-and-software bundles to help you get started quickly. However, since DVI is fundamentally an algorithm, you can also license it as a standalone software library for deployment on your own hardware. This standalone approach may be preferable if you need higher speeds than our bundled solutions can provide.
Starter Kit
Engicam carrier board for SMARC SoM with NXP i.MX 8M+, 2GB RAM, 8GB eMMC, Hailo-8 on M.2. 2x GbEth, 1x USB3, 2x USB2, LVDS, HDMI, RTC, etc.
Opt: 1x USB camera: global shutter, 100fps, color, 1MP, with HW trigger connected to carrier GPIOs.
XXX – board only
XXX – with camera
(distributed by us)
System on Module
SoM with NXP i.MX 8M+, with passive heatsink. Preinstalled Linux yocto with DVI APIs.
400€
(distributed by Avnet Silica as SilicaDVI01)
Industrial PC
Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum. Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.
XXX – board only
XXX – with camera
(distributed by us)
Smart Camera
SCAILX is 42mm x 42mm, PoW, 2MP global shutter smart camera, hosting an NXP i.MX 8M+. Preinstalled Linux yocto with DVI APIs.
850€
(distributed by Videology as ScailxDVI01)
Software library
Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum
XXX
(distributed by us)
In every case, DVI is licensed on a per-device basis, with no time limits. Training is included, and there are no recurring fees. Each license comes with a demo application, complete with open-source code for developers, and code snippets that demonstrate more advanced functionalities.