| Description |
Deep Learning (DL) and Convolutional Neural Networks (CNNs) have been established as a powerful class of models for image recognition problems. Encouraged by this, we implemented a system that can accurately predict handwritten numbers from a raw CMOS machine image. Current forms of deep learning do not target non-anthropomorphic camera images that our sensors will be producing. The concept of the project is that an implemented machine system can interpret our CMOS camera data thus having the ability to make out what it captures. To develop our system it took: creating a dataset, developing a trained convolutional neural network (CNN), and testing our system on live images. Our results included {0,1} at 99%, {0-4} at 80.6%, and {0-9} at 57.0% prediction accuracy respectively. In theory, from this we will have composed a system that can function on "non-human cameras as the eyes for the internet of things" for every machine system with an image capturing sensor embedded within it. |
