DogPoseDetectorOnnx.cpp
7.2 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
#include "DogPoseDetectorOnnx.h"
#include "logger.h"
#include "ErrorRecorder.h"
#include "logging.h"
#include "common.h"
#include "cuda_kernels.h"
#include <algorithm>
#include <fstream>
#include "opencv2/opencv.hpp"
using namespace nvinfer1;
SampleErrorRecorder gRecorder;
namespace sample
{
Logger gLogger{ Logger::Severity::kINFO };
LogStreamConsumer gLogVerbose{ LOG_VERBOSE(gLogger) };
LogStreamConsumer gLogInfo{ LOG_INFO(gLogger) };
LogStreamConsumer gLogWarning{ LOG_WARN(gLogger) };
LogStreamConsumer gLogError{ LOG_ERROR(gLogger) };
LogStreamConsumer gLogFatal{ LOG_FATAL(gLogger) };
void setReportableSeverity(Logger::Severity severity)
{
gLogger.setReportableSeverity(severity);
gLogVerbose.setReportableSeverity(severity);
gLogInfo.setReportableSeverity(severity);
gLogWarning.setReportableSeverity(severity);
gLogError.setReportableSeverity(severity);
gLogFatal.setReportableSeverity(severity);
}
} // namespace sample
DogPoseDetectorOnnx::DogPoseDetectorOnnx()
{
}
DogPoseDetectorOnnx::~DogPoseDetectorOnnx()
{
}
static size_t getFileSize(FILE* file) {
size_t fileSize = -1;
if (file != NULL) {
if (fseek(file, 0L, SEEK_END) == 0) {
fileSize = ftell(file);
}
rewind(file);
}
return fileSize;
}
bool DogPoseDetectorOnnx::init() {
const char* model_path_onnx = "../weights/best.onnx";
const char* input_node_name = "images";
const char* output_node_name = "output";
const char* plan_file_name = "dog_pose_detect_plan.bin";
FILE* f = fopen(plan_file_name, "rb");
if (f == nullptr)
{
auto builder = SampleUniquePtr<nvinfer1::IBuilder>(nvinfer1::createInferBuilder(sample::gLogger.getTRTLogger()));
if (!builder)
{
return false;
}
const auto explicitBatch = 1U << static_cast<uint32_t>(NetworkDefinitionCreationFlag::kEXPLICIT_BATCH);
auto network = SampleUniquePtr<nvinfer1::INetworkDefinition>(builder->createNetworkV2(explicitBatch));
if (!network)
{
return false;
}
auto config = SampleUniquePtr<nvinfer1::IBuilderConfig>(builder->createBuilderConfig());
if (!config)
{
return false;
}
auto parser = SampleUniquePtr<nvonnxparser::IParser>(nvonnxparser::createParser(*network, sample::gLogger.getTRTLogger()));
if (!parser)
{
return false;
}
auto parsed = parser->parseFromFile(model_path_onnx, static_cast<int>(sample::gLogger.getReportableSeverity()));
if (!parsed)
{
return false;
}
if (m_bUseFP16)
{
config->setFlag(BuilderFlag::kFP16);
}
else
{
config->setFlag(BuilderFlag::kINT8);
samplesCommon::setAllDynamicRanges(network.get(), 127.0F, 127.0F);
}
samplesCommon::enableDLA(builder.get(), config.get(), m_dlaCore);
// CUDA stream used for profiling by the builder.
auto profileStream = samplesCommon::makeCudaStream();
if (!profileStream)
{
return false;
}
config->setProfileStream(*profileStream);
SampleUniquePtr<IHostMemory> plan{ builder->buildSerializedNetwork(*network, *config) };
//plan = SampleUniquePtr<IHostMemory>(builder->buildSerializedNetwork(*network, *config));
if (!plan)
{
return false;
}
FILE* fp = fopen(plan_file_name, "wb");
fwrite(plan->data(), 1, plan->size(), fp);
fclose(fp);
mRuntime = std::shared_ptr<nvinfer1::IRuntime>(createInferRuntime(sample::gLogger.getTRTLogger()));
if (!mRuntime)
{
return false;
}
mEngine = std::shared_ptr<nvinfer1::ICudaEngine>( mRuntime->deserializeCudaEngine(plan->data(), plan->size()), samplesCommon::InferDeleter());
if (!mEngine)
{
return false;
}
}
else {
size_t file_size = getFileSize(f);
void* pPlanData = malloc(file_size);
fread(pPlanData, 1, file_size, f);
bool bSucceed = false;
do
{
mRuntime = std::shared_ptr<nvinfer1::IRuntime>(createInferRuntime(sample::gLogger.getTRTLogger()));
if (!mRuntime)
{
break;
}
mEngine = std::shared_ptr<nvinfer1::ICudaEngine>( mRuntime->deserializeCudaEngine(pPlanData, file_size), samplesCommon::InferDeleter());
if (!mEngine)
{
break;
}
bSucceed = true;
} while (0);
free(pPlanData);
if (!bSucceed) {
return false;
}
}
context = SampleUniquePtr<nvinfer1::IExecutionContext>(mEngine->createExecutionContext());
if (!context)
{
return false;
}
// 创建GPU显存缓冲区
m_data_buffer = new void*[2];
// 创建GPU显存输入缓冲区
m_input_node_index = mEngine->getBindingIndex(input_node_name);
m_input_node_dim = mEngine->getBindingDimensions(m_input_node_index);
size_t input_data_length = m_input_node_dim.d[1] * m_input_node_dim.d[2] * m_input_node_dim.d[3];
cudaMalloc(&(m_data_buffer[m_input_node_index]), input_data_length * sizeof(float));
// 创建GPU显存输出缓冲区
m_output_node_index = mEngine->getBindingIndex(output_node_name);
m_output_node_dim = mEngine->getBindingDimensions(m_output_node_index);
size_t output_data_length = m_output_node_dim.d[1] * m_output_node_dim.d[2];
cudaMalloc(&(m_data_buffer[m_output_node_index]), output_data_length * sizeof(float));
return true;
}
std::vector<DogPoseResult> DogPoseDetectorOnnx::detect(unsigned char *pGpuBgr, int src_width, int src_height) {
int dst_width = m_input_node_dim.d[2];
int dst_height = m_input_node_dim.d[3];
int max_side_length = std::max(src_width, src_height);
//saveCUDAImg(pGpuBgr, src_width, src_height, "src.jpg");
cudaStream_t stream;
cudaStreamCreate(&stream);
{
// 显存-->内存-->显存
//int rgb_size = 3 * src_width * src_height;
//uint8 *cpu_data = new uint8[rgb_size];
//cudaError_t cudaStatus = cudaMemcpy(cpu_data, pGpuBgr, rgb_size * sizeof(uint8), cudaMemcpyDeviceToHost);
//cv::Mat image(src_height, src_width, CV_8UC3, cpu_data);
//cv::Mat max_image = cv::Mat::zeros(cv::Size(max_side_length, max_side_length), CV_8UC3);
//cv::Rect roi(0, 0, image.cols, image.rows);
//image.copyTo(max_image(roi));
//// 将图像归一化,并放缩到指定大小
//cv::Size input_node_shape(m_input_node_dim.d[2], m_input_node_dim.d[3]);
//cv::Mat BN_image = cv::dnn::blobFromImage(max_image, 1 / 255.0, input_node_shape, cv::Scalar(0, 0, 0), true, false);
//size_t input_data_length = m_input_node_dim.d[1] * m_input_node_dim.d[2] * m_input_node_dim.d[3];
//std::vector<float> input_data(input_data_length);
//memcpy(input_data.data(), BN_image.ptr<float>(), input_data_length * sizeof(float));
//cudaMemcpyAsync(m_data_buffer[m_input_node_index], input_data.data(), input_data_length * sizeof(float), cudaMemcpyHostToDevice, stream);
}
cuda_common::resizeAndNorm(pGpuBgr, src_width, src_height, (float*)m_data_buffer[m_input_node_index], dst_width, dst_height);
// 模型推理
context->enqueueV2(m_data_buffer, stream, nullptr);
size_t output_data_length = m_output_node_dim.d[1] * m_output_node_dim.d[2];
float* result_array = new float[output_data_length];
cudaMemcpyAsync(result_array, m_data_buffer[m_output_node_index], output_data_length * sizeof(float), cudaMemcpyDeviceToHost, stream);
cudaDeviceSynchronize();
ResultYolov5 result;
result.factor = max_side_length / (float)m_input_node_dim.d[2];
result.read_class_names("../weights/classes.txt");
return result.yolov5_result(result_array, 0.6);
}