A Computer Vision Pipeline to Match Lost and Found Dogs
By Anant Sunilam Awasthy, Rishabh Kaushal, Karthik Srinatha, Aidan Vickars
In this project we developed an Android application where users can submit an image of their lost dog, and the most similar lost dogs that have been found will be shown. Similarly, users that find a lost dog will submit an image, and the most similar dogs that have been lost will be returned. The app uses three convolutional neural networks that process visual input, analyze dog breeds, and eventually calculate a similarity score between two images to identify if there is a match. We determined that a successful result is achieved if the lost dog is included in the top 15 found dogs. Our application delivered this outcome 89% of the time.