Micaela Verucchi
35 lines
1.1 KiB
C++
35 lines
1.1 KiB
C++
#include<iostream>
|
|
#include<vector>
|
|
#include "tkdnn.h"
|
|
#include "test.h"
|
|
#include "DarknetParser.h"
|
|
|
|
int main() {
|
|
std::string bin_path = "yolo4_416";
|
|
std::vector<std::string> input_bins = {
|
|
bin_path + "/layers/input.bin"
|
|
};
|
|
std::vector<std::string> output_bins = {
|
|
bin_path + "/debug/layer139_out.bin",
|
|
bin_path + "/debug/layer150_out.bin",
|
|
bin_path + "/debug/layer161_out.bin"
|
|
};
|
|
std::string wgs_path = bin_path + "/layers";
|
|
std::string cfg_path = "../tests/darknet/cfg/yolo4_416.cfg";
|
|
std::string name_path = "../tests/darknet/names/coco.names";
|
|
downloadWeightsifDoNotExist(input_bins[0], bin_path, "https://cloud.hipert.unimore.it/s/982LxTQcNQfFQc4/download");
|
|
|
|
// parse darknet network
|
|
tk::dnn::Network *net = tk::dnn::darknetParser(cfg_path, wgs_path, name_path);
|
|
net->print();
|
|
|
|
//convert network to tensorRT
|
|
tk::dnn::NetworkRT *netRT = new tk::dnn::NetworkRT(net, net->getNetworkRTName(bin_path.c_str()));
|
|
|
|
int ret = testInference(input_bins, output_bins, net, netRT);
|
|
net->releaseLayers();
|
|
delete net;
|
|
delete netRT;
|
|
return ret;
|
|
}
|