Speaker
Description
In the context of scientific computing there is a growing interest towards DNNs (Deep Neural Networks) increasingly used in several scientific applications, spanning from medical imaging segmentation and classification, to on-line analysis of experimental data.
Despite GPUs are the most used accelerators in this context, FPGAs are also emerging as a "new" technology promising higher energy-efficiency and lower latency solutions, in particular for the inference phase of DNNs and ML algorithms.
In this talk, after a brief introduction on FPGA devices and programming frameworks, the results for the inference phase of an actual application, concerning the identification of calcium lesions on contrast enhanced CT-scans, will be presented, showing a comparison with GPU accelerators in terms of time-to-solution, accuracy and energy-efficiency.
