An expert in vision systems, he speaks from experience, having developed control systems in very diverse industries: "In the food industry, the product is dynamic; it is rarely 100% identical. When it is inspected by humans, the conformity decision is the result of consultation between different experts, on different types of defects. The challenge for the vision system is to take into account all these areas of expertise and, at the same time, to smooth out human subjectivity."
Baptiste Amato-Gagnon recommends a pragmatic and progressive approach to ensuring the performance of automated quality systems with vision because, as he says, "there's no magic in technology." For him, the only systems capable of meeting the needs of the food industry are learning and supervised systems: "This allows the machine to eliminate criteria that are not relevant for characterizing a particular type of defect. As a result, the system will be much more precise. I've seen unsupervised systems with enormous rejection rates, which are not at all representative of the actual quality." A learning system is also a system that allows the company to capitalize on human expertise. "When an operator removes a product from the line, they contribute their value at a specific moment." “If the operator teaches the machine why it did something, it creates value that the company can leverage for years,” explains Baptiste Amato-Gagnon, who adds: “The machine will also return this value to the operator by allowing them to more easily identify the root causes of their production problems, enabling them to return to a compliant product more quickly.” To food manufacturers who might still be hesitant to adopt this approach, Baptiste Amato-Gagnon concludes with one final piece of advice: “Don’t try to identify all the defects right away. It’s better to focus on the defects that are costing the most money. This way, you can achieve good results very quickly, with a return on investment in just a few months, or even a few weeks.”
