Systolic-based 2D convolver for CNN in FPGA

Convolution is a primary mathematical operation used in many signal processing and analysis algorithms. High dependence of the complex systems on the correct operation of the convolver demands its continual improvements mostly related to the decrease of resource consumption. The paper proposes a model of 2D convolution massively used in the algorithms of image processing. The paper provides a detailed description of the model structure with focus on the implementation aspect. The model is particularly applied to the convolutional layer of Convolutional Neural Network, currently the most known image-based deep learning method. The key difference of the proposed model compared with other common implementations lies in the placement of line buffers. The correctness of the model design is validated through the simulation discussed at the end of paper.