Prototype and Adjust a Deep Learning Network on FPGA

Deep learning inferencing is so computationally intensive that using it in FPGA-based applications often requires significant customization of the hardware architecture and the deep learning network itself. Being able to iterate during early exploration is vital to converging on your goals during implementation.

Deep Learning HDL Toolbox™ works closely with Deep Learning Toolbox™ within MATLAB® to let deep learning practitioners and hardware designers prototype and make large-scale changes to meet system requirements. This video shows how to:

• Prototype a pretrained industrial defect detection network on an FPGA

• Analyze network performance running on the FPGA

• Adjust the network design and quickly prototype to see the results

• Quantize the network and parameters to int8 data types