This post might be too technical for some, but if you're willing to wade in, it's not too technical, and gives you an appreciation for one particular aspect of artificial intelligence: object detection. One major type of AI used for this is the the convolutional neural network (CNN). These are composed of convolutional 'layers', each of which extracts a specific feature from (say) an image; these are then 'pooled' to reduce the set of feature maps and zero in on an object. A more recent type of AI is the the region-based convolutional neural network (R-CNN), and even more recently, Fast R-CNN and Faster R-CNN which bring region selection and feature selection into a single model. Finally, there's "You Only Look Once" (YOLO), which "instead of extracting features and regions separately, performs everything in a single pass through a single network."
Today: 7 Total: 102 [Direct link] [Share]
