The integration of artificial intelligence (AI) in various sectors such as industrial, medical and aerospace, is becoming increasingly prevalent. However, the question of how to get started with AI is not as important as determining what kind of AI is best suited for a specific application. Avnet, a leading technology solutions provider, understands that their customers are primarily concerned with results and by understanding the use case, the best solution for AI often presents itself.
One area where Avnet can offer significant value is in helping software companies with computer vision experience move into new use cases. According to Michaël Uyttersprot, Manager of System Solutions, AI/ML & Vision, EMEA with Avnet Silica, “Part of my role is to ensure we have the expertise and partners needed to help companies with computer vision experience move into these new use cases.”
When it comes to developing a Machine vision system, selecting the right image sensor is crucial, especially when Machine Learning (ML) is going to be used to analyze the image data. For example, if the goal is to use machine vision with ML to check something fast-moving, like a bottle on a production line, a camera module with a global shutter is required to capture all the necessary details. Additionally, the image sensor needs to be appropriate for the operating environment, such as having the appropriate low-light performance.
Sensor fusion, the process of combining data from multiple sources into a single set that can be fed into a neural network, is also crucial for the success of machine learning in industry. However, there is no standard way to do this, which means it is up to the manufacturer to handle the sensor fusion.
Creating a demonstrator to prove the concept of machine learning is not as difficult as it used to be, thanks to the availability of evaluation modules and development kits from distribution partners. However, taking the concept to production is still a challenge. Techniques such as transfer learning can be useful, which takes data created for one use case and ports it across to a new but similar use-case. Another solution is to create synthetic data virtually using 3D modeling,
which allows many images to be created quickly and used to train a machine learning algorithm without the effort or expense of physically creating examples. The question of whether to use edge or cloud processing for machine learning is becoming more prevalent as AI is moving to the network’s edge. However, Uyttersprot points out that both edge and cloud processing can coexist and offer benefits that the other cannot.
The post Commerce Machine Learning first appeared on Mega Components.
The post Commerce Machine Learning appeared first on Mega Components.
