Cyril Poulet

About

Cyril Poulet is a pioneering researcher in embedded computer vision and hardware-accelerated deep learning, best known for his foundational work on FPGA-based processors for convolutional neural networks. His landmark 2009 paper, "CNP: An FPGA-based processor for Convolutional Networks," has garnered over 360 citations, establishing him as a key figure in bridging efficient neural network inference with reconfigurable hardware. Poulet’s research focuses on real-time, low-power vision systems, where he demonstrated that convolutional networks—biologically inspired hierarchical architectures for detection, recognition, and segmentation—can be effectively deployed on FPGAs for embedded applications. His follow-up work on an FPGA-based stream processor for real-time vision further advanced this field, enabling practical deployment of multi-layer convolutional filter banks and non-linearities in resource-constrained environments. By addressing the computational demands of ConvNets, HMAX-type architectures, and dense SIFT features, Poulet’s contributions have directly influenced the development of efficient edge-AI systems. His achievements are particularly notable for their impact on autonomous robotics and surveillance, where real-time performance is critical. For students and researchers, Poulet’s work exemplifies how hardware-software co-design can unlock the potential of deep learning outside the data center.

Research Focus

Key Achievements

Top Papers

1. 1
   CNP: An FPGA-based processor for Convolutional Networks

   362 citations · 2009
2. 2
   An FPGA-based stream processor for embedded real-time vision with Convolutional Networks

   54 citations · 2009

Key Collaborators

Contact & Links