About this document
CNN Accelerator Hardware Dataflow by ofdm.pme is a document available to read on EtoBox.
This paper presents a novel hardware dataflow for Convolutional Neural Network (CNN) accelerators, focusing on energy efficiency and scalability for edge computing applications. It proposes an input stationary dataflow that optimizes the reuse of input activations and can be implemented on FPGA or ASIC, achieving 30 GOPS throughput. The research also includes an analytical framework for estimating energy consumption and data reusability under various workloads.
- Author
- ofdm.pme
- Language
- EN