to illustrate how to removal a Neural Network from pc to FPGA board ,it contain all the code include c code worked in pc,HLS prj accelerate code in FPGA,and implacement in block design and with SDK app......
the Neural Network c code ,you can train the nnet and test in this proj by pc
cd cnet/mnist
make -----to generate the libnet.a Library file
make train --- to generate the elf file in linux system and train the nnet in PC and
deploy with ./train ,you can get the weight file in ./mnist/out until the train over ,,it contain both pc file and FPGA file which can be used by hls format with ".dat",you can recognise them by differernt name
make test--- to generate the elf file in linux syetem and test the nnet's accuracy and show
and deploy with ./train in linux
make save_data ---to generate the data file from orign file to format ".dat"
deployed ways ./save_data,you can modefythe number in source file to change the number of files which can be used in HLS proj or used in SDK proj ,,they named "id_true.dat".
the project code to generate the HLS ip core which can be used in FPGA
the project code contain the block design and the .bit file and .hdf ,
you can reference to csdn link:https://blog.csdn.net/u012116328/article/details/117246023
from SDU CNSATM TEAM!