A trained Convolutional Neural Network implemented on ZedBoard Zynq-7000 FPGA.

A real-time SLR system combining CNN-LSTM and CNN-Transformer architectures. Trained on a custom dataset, the models extract spatial features and capture temporal dependencies for accurate gesture recognition. Includes full source code for data preprocessing, training, and real-time inference.

This is an excerpt from the project report of the final year project named: Deep Learning (SLR) on FPGA, the part pertaining to the summary of the literature review of the techniques used for the optimization and acceleration of the deep learning especially the convolutional neural networks (CNN) on edge devices, embedded systems, FPGA, cloud etc.