Bookmark and Share

ARM Demonstration of Collaboratively Optimizing Deep Learning Applications

Gian Marco Iodice, Software Engineer at ARM, demonstrates the company's latest embedded vision technologies and products at the May 2017 Embedded Vision Summit. Specifically, Iodice presents a demonstration of Collective Knowledge, developed by partner company Dividiti. The demo supports several popular models for object classification and detection, implemented in Caffe or TensorFlow, and running on the CPU or on the GPU via OpenCL. The user can observe how different choices affect the speed and accuracy of recognition on several popular datasets, including ImageNet and KITTI.