Computational and Communication Reduction Technique in Machine Learning Based Near Sensor Applications

Mohamed Ayoub Neggaz (1) , Smail Niar (1, 2) , Fadi Kurdahi (3)

1 LAMIH - Laboratoire d'Automatique, de Mécanique et d'Informatique industrielles et Humaines - UMR 8201
2 INSA Hauts-De-France - INSA Institut National des Sciences Appliquées Hauts-de-France
3 CECS - Center for Embedded Computer Systems [Irvine]

State-of-the-art Convolutional Neural Networks (CNN) are used to process images. In most cases, videos are streamed and processed frame by frame using a CNN. In this paper we present a two-step approach to process images in a real-life streaming environment. We exploit size-reduction and data encoding to reduce the computational and communication load. A near-sensor architecture is proposed. The final design reaches 14 EPS for the full Faster R-CNN pipeline.

Domaines: Informatique [cs]

Mohamed Ayoub Neggaz, Smail Niar, Fadi Kurdahi. Computational and Communication Reduction Technique in Machine Learning Based Near Sensor Applications. 2018 30th International Conference on Microelectronics (ICM), Dec 2018, Sousse, Tunisia. pp.68-71, ⟨10.1109/ICM.2018.8704033⟩. ⟨hal-03382375⟩