EZSort
Abstract — Organization is something that everyone often wishes to attain but can be tedious and cumbersome to achieve. Having to manually sift through hundreds of items and separate them into groups can be time-consuming and tiresome. We have created a modular item sorter that can be configured to any set of items rather than being specialized to just one use case. By interfacing with the device, the user can easily train, configure, and use the sorter to neatly sort between three groups while also accounting for foreign objects. The device makes use of the MobileNet convolutional neural network (CNN) and the k-nearest neighbors (KNN) algorithm to create an efficient method of identifying, classifying, and grouping items based on features. We believe this method can be expanded to become even more efficient thus alleviating the need for separate specialized machinery in the field of sorting.