Machine learning has come a far way in the last 5-10 years especially when it comes to edge applications. This is true as two main trends are manifesting a) more and more cloud based compute is being shifted to the edge and more importantly b) the edge is becoming capable of doing more complex tasks than ever before. The main drivers for these advancements are to be found in machine learning on the edge and hardware improvements. With this in mind we will share with you how we deploy transfer learning to create cutting edge solutions like few shot key word spotting and a novel architecture for object detection on MCUs - faster objects, more objects FOMO.

Download the slides for this talk.Download ( PDF, 2772.35 MB)