Implémentation Efficiente de Fonctions de Convolution sur FPGA à l'Aide de Blocs Paramétrables et d'Approximations Polynomiales
This work addresses the challenge of optimizing CNN deployment on FPGAs for applications requiring low latency and power efficiency, though it is incremental in improving existing hardware implementation methods.
The paper tackles the complexity of implementing convolutional neural networks (CNNs) on FPGAs by proposing configurable convolution blocks and a framework for predicting resource utilization, showing that the blocks adapt to hardware constraints and the models accurately predict resource consumption.
Implementing convolutional neural networks (CNNs) on field-programmable gate arrays (FPGAs) has emerged as a promising alternative to GPUs, offering lower latency, greater power efficiency and greater flexibility. However, this development remains complex due to the hardware knowledge required and the long synthesis, placement and routing stages, which slow down design cycles and prevent rapid exploration of network configurations, making resource optimisation under severe constraints particularly challenging. This paper proposes a library of configurable convolution Blocks designed to optimize FPGA implementation and adapt to available resources. It also presents a methodological framework for developing mathematical models that predict FPGA resources utilization. The approach is validated by analyzing the correlation between the parameters, followed by error metrics. The results show that the designed blocks enable adaptation of convolution layers to hardware constraints, and that the models accurately predict resource consumption, providing a useful tool for FPGA selection and optimized CNN deployment.