Imperial College London > Talks@ee.imperial > CAS Talks > A Framework for Mapping Multiple Convolutional Neural Networks on FPGAs

A Framework for Mapping Multiple Convolutional Neural Networks on FPGAs

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact George A Constantinides.

Since 2012, Convolutional Neural Networks (CNNs) have demonstrated state-of-the-art performance in numerous Artificial Intelligence (AI) tasks. This property has made CNNs an enabling technology for various emerging AI systems, from self-driving cars to large-scale image captioning. Such systems often employ multiple CNN models, with each one trained for a particular task of the system. In such systems, the competition for resources by the multiple CNNs may lead to suboptimal performance. In this context, the efficient mapping of multiple CNNs on the system’s computing platforms that respects the application-level performance and power requirements is essential. At the same time, FPG As have emerged as an attractive platform for computatinally intensive CNN -based applications due to their customisation and reconfigurability capabilities that are combined with low power consumption. In this talk, we will address the problem of the optimised mapping of multiple CNNs on FPG As in the context of multi-CNN systems. A framework will be presented that consists of a novel multi-CNN hardware architecture together with an automated design methodology for the configuration of the architecture based on the application-level performance requierments for each CNN .

This talk is part of the CAS Talks series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


Changes to Talks@imperial | Privacy and Publicity