hi all, at present, there is only C + + classification code about TVM deployment, but there is no target detection code. If anyone sees this problem, please help me solve the problem and share your TVM version code about C + + Darknet. Thank you very much!!!
Here is my code, but I can only write so much, I can’t realize the following processing
#include <dlpack/dlpack.h>
#include <tvm/runtime/module.h>
#include <tvm/runtime/registry.h>
#include <tvm/runtime/packed_func.h>
#include <fstream>
#include <opencv2/opencv.hpp>
#include <cuda_runtime.h>
using namespace std;
void Mat_to_CHW(float *data, cv::Mat &frame)
{
assert(data && !frame.empty());
unsigned int volChl = 416 * 416;
for(int c = 0; c < 3; ++c)
{
for (unsigned j = 0; j < volChl; ++j)
data[c*volChl + j] = static_cast<float>(float(frame.data[j * 3 + c]) / 255.0);
}
}
int main() {
tvm::runtime::Module mod_dylib = tvm::runtime::Module::LoadFromFile("/data_2/git/onnx_to_tvm/darknet.so");
std::ifstream json_in("/data_2/git/onnx_to_tvm/darknet.json", std::ios::in);
std::string json_data((std::istreambuf_iterator<char>(json_in)), std::istreambuf_iterator<char>());
json_in.close();
// parameters in binary
std::ifstream params_in("/data_2/git/onnx_to_tvm/darknet.params", std::ios::binary);
std::string params_data((std::istreambuf_iterator<char>(params_in)), std::istreambuf_iterator<char>());
params_in.close();
TVMByteArray params_arr;
params_arr.data = params_data.c_str();
params_arr.size = params_data.length();
int dtype_code = kDLFloat;
int dtype_bits = 32;
int dtype_lanes = 1;
int device_type = kDLGPU;
int device_id = 0;
tvm::runtime::Module mod = (*tvm::runtime::Registry::Get("tvm.graph_runtime.create"))
(json_data, mod_dylib, device_type, device_id);
DLTensor *x;
int in_ndim = 4;
int64_t in_shape[4] = {1, 3, 416, 416};
TVMArrayAlloc(in_shape, in_ndim, dtype_code, dtype_bits, dtype_lanes, device_type, device_id, &x);
//其中DLTensor是个灵活的结构，可以包容各种类型的张量，而在创建了这个张量后，我们需要将OpenCV中读取的图像信息传入到这个张量结构中：
// 这里依然读取了papar.png这张图
cv::Mat image = cv::imread("/home/em/.tvm_test_data/data/dog.jpg");
cv::Mat frame, input;
cv::cvtColor(image, frame, cv::COLOR_BGR2RGB);
cv::resize(frame, input, cv::Size(416,416), cv::INTER_CUBIC);
float data[416 * 416 * 3];
// 在这个函数中 将OpenCV中的图像数据转化为CHW的形式
Mat_to_CHW(data, input);
cudaMemcpy(x->data, &data, 3 * 416 * 416 * sizeof(float), cudaMemcpyHostToDevice);
//memcpy(x->data, &data, 3 * 416 * 416 * sizeof(float));
// get the function from the module(set input data)
tvm::runtime::PackedFunc set_input = mod.GetFunction("set_input");
set_input("data", x);
// get the function from the module(load patameters)
tvm::runtime::PackedFunc load_params = mod.GetFunction("load_params");
load_params(params_arr);
DLTensor* y;
int out_ndim = 1;
int64_t out_shape[4] = {1, 255 ,52, 52};
TVMArrayAlloc(out_shape, out_ndim, dtype_code, dtype_bits, dtype_lanes, device_type, device_id, &y);
// get the function from the module(run it)
tvm::runtime::PackedFunc run = mod.GetFunction("run");
// get the function from the module(get output data)
tvm::runtime::PackedFunc get_output = mod.GetFunction("get_output");
run();
tvm::runtime::NDArray res;
for(int k = 0; k< 3 ; ++k){
std::cout << "yes" << std::endl;
res = get_output(k*4);
float p_res[255*52*52];
cudaMemcpy(p_res, res->data, 255 * 52 * 52 * sizeof(float), cudaMemcpyDeviceToHost);
//std::cout << res->ndim << std::endl;
std::cout << res.Shape()[0] << std::endl;
std::cout << res.Shape()[1] << std::endl;
std::cout << res.Shape()[2] << std::endl;
std::cout << res.Shape()[3] << std::endl;
// cv::Mat out(int(res.Shape()[2]),int(res.Shape()[3]),CV_32FC(255));
// cudaMemcpy(out.data, res->data, int(res.Shape()[2])*int(res.Shape()[3])*255* sizeof(float), cudaMemcpyDeviceToHost);
// cout << out.cols << out.rows << out.channels() << endl;
// std::cout << out << std::endl;
}
return 0;
}