Visual quality control in the food industry with Edge AI

Problem Statement: Food producers strive for consistent product quality, but human visual inspection often falls short. At high production speeds, human inspectors can become fatigued and overlook minor discrepancies. This leads to the risk that products with defects or contaminations end up at the customer's door, resulting in costly recalls - on average, a single product recall costs large companies as much as $10 million. Moreover, 24/7 manual inspection is not feasible, leaving gaps in quality control.

Image: Rugged IP66/67 panel PCs with integrated NVIDIA Jetson modules offer up to 100 TOPS of AI computing power for visual inspection in humid, hygienic environments. In practice, one places a SimplexWaterproof panel PC (fully IP67 waterproof, food-grade stainless steel housing, passively cooled) right next to the production line. This industrial computer with, for example, an NVIDIA Jetson module or AMD Ryzen™ AI accelerator performs inference locally. Via high-speed cameras, the system continuously receives images of passing products and analyzes on the spot whether each item meets quality standards. If discrepancies are detected, the panel PC immediately sends a signal to a handler or alarm without delay.

Advantages over cloud: A local Edge AI approach eliminates the latency of cloud-based vision assessment: processing happens on-site without significant delay, essential in fast-moving production lines. In addition, sensitive production data (such as camera images of products) stays indoors - this increases data security and safeguards customer-specific recipe or process information. Also, the solution is independent of Internet connectivity; thus, production does not stop in the event of network outages. Finally, scaling up is easy by adding an additional panel PC per line, instead of sending bandwidth-intensive video streams continuously to the cloud.