First, I convert the models of yolo v3 to json
from gluoncv import model_zoo, data, utils
net = model_zoo.get_model('yolo3_darknet53_coco', pretrained=True)
im_fname = utils.download('https://raw.githubusercontent.com/zhreshold/' +
'mxnet-ssd/master/data/demo/dog.jpg',
path='dog.jpg')
x, img = data.transforms.presets.yolo.load_test(im_fname, short=320)
print('Shape of pre-processed image:', x.shape)
net.hybridize()
class_IDs, scores, bounding_boxs = net(x)
net.export("yolov3.json", epoch=279)
Second, load the yolov3.json by c++ api
#include <mxnet-cpp/MxNetCpp.h>
#include <iostream>
using namespace mxnet::cpp;
using namespace std;
int main()try
{
std::string const prefix("C:/Users/yyyy/.mxnet/models/yolo3");
Symbol::Load(prefix + ".json");
return 0;
}catch(std::exception const &ex){
cerr<<ex.what()<<endl;
}
But it throw exception
[21:09:34] …\3rdLibs\incubator-mxnet\build\install\include\mxnet-cpp/symbol.hpp:102: Check failed: MXSymbolCreateFromFile(file_name.c_str(), &(handle)) == 0 (-1 vs. 0)
What kind of error I commit?How could I solve it?Thanks
ps : This codes can load the symbol from MTCNN.