Real-Time Object Recognition on Mobile

Object recognition using convolutional neural networks has been extensively used in the cloud because of its computing costs. However the growth of lightweight hardware like drones and mobile augmented reality turned people's attention towards faster, local solutions that would implement deep learning directly into connected devices. From security cameras that know what they see, to autonomous delivery robots the size of small suitcases, to wearables and apps that recognize faces and objects faster than most humans are capable of, deep learning on mobile hardware is set to change the way we think of apps, IoT, and robotics.

We will be looking into present uses of artificial intelligence for vision used locally, and discuss the implications of this emerging trend for app development, and IoT products, as well as situation where this configuration fails. In particular, we will explore the way mobile object recognition has had an impact in the accessibility sector in helping the blind recognize their surroundings, and personal learnings from working with visually impaired users who depended on the performance of an AI technology from a user experience and human-computer interaction standpoint.

Speaker: Alberto Rizzoli, Aipoly