Available On Demand
It is true that you can’t judge the taste of an apple by its appearance. However, a single bruised apple emits ethylene, which can dramatically reduce the shelf life of the surrounding fruit. The result is that the agricultural supply-chain poses an interesting and ever more pressing challenge for machine vision. If the apple is better suited for mom’s apple pie, or grandma’s apple sauce, it can be accurately sorted for the same, while ensuring the longevity of the fruit that are destined for the produce aisle in your local supermarket.
Is this concept new? Not really. The first electro-optical produce sorting machines came to market in the 1930s, and the field of classical computer vision can be traced back to the 1950s. Traditional computer vision gave us key techniques such as Edge Detection, Optical Flow, SIFT and SURF that continue to provide optimum performance in many applications. Moreover, it is also true that convolution kernel weights were hand-crafted to extract important visual features long before training CNNs became all the rage.
However, the use of adaptable hardware for acceleration of these functions is value proposition that remains uniquely Xilinx. While many applications today leverage deep-learning techniques such as segmentation, classification and detection, there is still a very important place for these classical vision algorithms. Join us to learn how Kria SOMs and the Vitis Vision Libraries can help you accelerate and augment edge intelligence with an adaptable and accelerated spin on classical computer vision.
AI System Architect, Xilinx