-
Notifications
You must be signed in to change notification settings - Fork 212
/
yolov8_seg.cpp
224 lines (204 loc) · 10.5 KB
/
yolov8_seg.cpp
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
#include"yolov8_seg.h"
#include"decode_yolov8_seg.h"
YOLOv8Seg::YOLOv8Seg(const utils::InitParameter& param) :yolo::YOLO(param)
{
m_output_objects_device = nullptr;
m_output_objects_width = 39;
m_output_src_width = 116;
m_output_obj_area = 1 + param.topK * m_output_objects_width;
m_output_seg_w = 160 * 160;
m_output_seg_h = 32;
int output_objects_size = param.batch_size * m_output_obj_area;
CHECK(cudaMalloc(&m_output_objects_device, output_objects_size * sizeof(float)));
m_output_objects_host = new float[output_objects_size];
m_mask160 = cv::Mat::zeros(1, 160 * 160, CV_32F);
m_mask_eigen160 = Eigen::MatrixXf(1, 160 * 160);
m_thresh_roi160 = cv::Rect(0, 0, 160, 160);
m_thresh_roisrc = cv::Rect(0, 0, m_param.src_w, m_param.src_h);
m_downsample_scale = 160.f / 640;
m_mask_src = cv::Mat::zeros(m_param.src_h, m_param.src_w, CV_32F);
m_img_canvas = cv::Mat::zeros(cv::Size(m_param.src_w, m_param.src_h), CV_8UC3);
}
YOLOv8Seg::~YOLOv8Seg()
{
CHECK(cudaFree(m_output_objects_device));
CHECK(cudaFree(m_output_src_device));
CHECK(cudaFree(m_output_src_transpose_device));
CHECK(cudaFree(m_output_seg_device));
delete[] m_output_objects_host;
delete[] m_output_seg_host;
m_output_src_device = nullptr;
}
bool YOLOv8Seg::init(const std::vector<unsigned char>& trtFile)
{
if (trtFile.empty())
{
return false;
}
std::unique_ptr<nvinfer1::IRuntime> runtime =
std::unique_ptr<nvinfer1::IRuntime>(nvinfer1::createInferRuntime(sample::gLogger.getTRTLogger()));
if (runtime == nullptr)
{
return false;
}
this->m_engine = std::unique_ptr<nvinfer1::ICudaEngine>(runtime->deserializeCudaEngine(trtFile.data(), trtFile.size()));
if (this->m_engine == nullptr)
{
return false;
}
this->m_context = std::unique_ptr<nvinfer1::IExecutionContext>(this->m_engine->createExecutionContext());
if (this->m_context == nullptr)
{
return false;
}
if (m_param.dynamic_batch)
{
this->m_context->setBindingDimensions(0, nvinfer1::Dims4(m_param.batch_size, 3, m_param.dst_h, m_param.dst_w));
}
m_output_dims = this->m_context->getBindingDimensions(2);
m_total_objects = m_output_dims.d[2];
assert(m_param.batch_size <= m_output_dims.d[0]);
m_output_area = 1; // 116 * 8400
for (int i = 1; i < m_output_dims.nbDims; i++)
{
if (m_output_dims.d[i] != 0)
{
m_output_area *= m_output_dims.d[i];
}
}
m_output_seg_dims = this->m_context->getBindingDimensions(1);
assert(m_param.batch_size <= m_output_seg_dims.d[0]);
m_output_seg_area = 1; // 32 * 160 * 160
for (int i = 1; i < m_output_seg_dims.nbDims; i++)
{
if (m_output_seg_dims.d[i] != 0)
{
m_output_seg_area *= m_output_seg_dims.d[i];
}
}
CHECK(cudaMalloc(&m_output_src_device, m_param.batch_size * m_output_area * sizeof(float)));
CHECK(cudaMalloc(&m_output_src_transpose_device, m_param.batch_size * m_output_area * sizeof(float)));
CHECK(cudaMalloc(&m_output_seg_device, m_param.batch_size * m_output_seg_area * sizeof(float)));
m_output_seg_host = new float[m_param.batch_size * m_output_seg_area];
float a = float(m_param.dst_h) / m_param.src_h;
float b = float(m_param.dst_w) / m_param.src_w;
float scale = a < b ? a : b;
cv::Mat src2dst = (cv::Mat_<float>(2, 3) << scale, 0.f, (-scale * m_param.src_w + m_param.dst_w + scale - 1) * 0.5,
0.f, scale, (-scale * m_param.src_h + m_param.dst_h + scale - 1) * 0.5);
cv::Mat dst2src = cv::Mat::zeros(2, 3, CV_32FC1);
cv::invertAffineTransform(src2dst, dst2src);
m_dst2src.v0 = dst2src.ptr<float>(0)[0];
m_dst2src.v1 = dst2src.ptr<float>(0)[1];
m_dst2src.v2 = dst2src.ptr<float>(0)[2];
m_dst2src.v3 = dst2src.ptr<float>(1)[0];
m_dst2src.v4 = dst2src.ptr<float>(1)[1];
m_dst2src.v5 = dst2src.ptr<float>(1)[2];
return true;
}
void YOLOv8Seg::preprocess(const std::vector<cv::Mat>& imgsBatch)
{
resizeDevice(m_param.batch_size, m_input_src_device, m_param.src_w, m_param.src_h,
m_input_resize_device, m_param.dst_w, m_param.dst_h, 114, m_dst2src);
bgr2rgbDevice(m_param.batch_size, m_input_resize_device, m_param.dst_w, m_param.dst_h,
m_input_rgb_device, m_param.dst_w, m_param.dst_h);
normDevice(m_param.batch_size, m_input_rgb_device, m_param.dst_w, m_param.dst_h,
m_input_norm_device, m_param.dst_w, m_param.dst_h, m_param);
hwc2chwDevice(m_param.batch_size, m_input_norm_device, m_param.dst_w, m_param.dst_h,
m_input_hwc_device, m_param.dst_w, m_param.dst_h);
}
bool YOLOv8Seg::infer()
{
float* bindings[] = { m_input_hwc_device, m_output_seg_device, m_output_src_device };
bool context = m_context->executeV2((void**)bindings);
return context;
}
void YOLOv8Seg::postprocess(const std::vector<cv::Mat>& imgsBatch)
{
yolov8seg::transposeDevice(m_param, m_output_src_device, m_total_objects, m_output_src_width, m_total_objects * m_output_src_width,
m_output_src_transpose_device, m_output_src_width, m_total_objects);
yolov8seg::decodeDevice(m_param, m_output_src_transpose_device, m_output_src_width, m_total_objects, m_output_area,
m_output_objects_device, m_output_objects_width, m_param.topK);
nmsDeviceV1(m_param, m_output_objects_device, m_output_objects_width, m_param.topK, m_output_obj_area);
CHECK(cudaMemcpy(m_output_objects_host, m_output_objects_device, m_param.batch_size * sizeof(float) * m_output_obj_area, cudaMemcpyDeviceToHost));
CHECK(cudaMemcpy(m_output_seg_host, m_output_seg_device, m_param.batch_size * sizeof(float) * m_output_seg_area, cudaMemcpyDeviceToHost));
}
void YOLOv8Seg::reset()
{
CHECK(cudaMemset(m_output_objects_device, 0, sizeof(float) * m_param.batch_size * m_output_obj_area));
}
void YOLOv8Seg::showAndSave(const std::vector<std::string>& classNames, const int& cvDelayTime, std::vector<cv::Mat>& imgsBatch)
{
cv::Point bbox_points[1][4];
const cv::Point* bbox_point0[1] = { bbox_points[0] };
int num_points[] = { 4 };
for (size_t bi = 0; bi < imgsBatch.size(); bi++)
{
int num_boxes = std::min((int)(m_output_objects_host + bi * m_output_obj_area)[0], m_param.topK);
float* p_seg = m_output_seg_host + bi * m_output_seg_area;
Eigen::Map<Eigen::MatrixXf> img_seg_(p_seg, m_output_seg_w, m_output_seg_h);
int m_output_obj_area_bi = bi * m_output_obj_area;
m_img_canvas.setTo(cv::Scalar(0, 0, 0));
for (size_t i = 0; i < num_boxes; i++)
{
float* ptr = m_output_objects_host + m_output_obj_area_bi + m_output_objects_width * i + 1;
if (ptr[6])
{
int label = ptr[5];
cv::Scalar color = utils::Colors::color80[label];
Eigen::Map<Eigen::MatrixXf> img_ojb_seg_(ptr + 7, m_output_seg_h, 1);
m_mask_eigen160 = img_seg_ * img_ojb_seg_;
cv::eigen2cv(m_mask_eigen160, m_mask160);
cv::exp(-m_mask160, m_mask160);
m_mask160 = 1.f / (1.f + m_mask160);
m_mask160 = m_mask160.reshape(1, 160);
int x_lt_160 = std::round(ptr[0] * m_downsample_scale);
int y_lt_160 = std::round(ptr[1] * m_downsample_scale);
int x_rb_160 = std::round(ptr[2] * m_downsample_scale);
int y_rb_160 = std::round(ptr[3] * m_downsample_scale);
cv::Rect roi160 = cv::Rect(x_lt_160, y_lt_160, x_rb_160 - x_lt_160, y_rb_160 - y_lt_160) & m_thresh_roi160;
if (roi160.width == 0 || roi160.height == 0)
continue;
int x_lt_src = std::round(m_dst2src.v0 * ptr[0] + m_dst2src.v1 * ptr[1] + m_dst2src.v2);
int y_lt_src = std::round(m_dst2src.v3 * ptr[0] + m_dst2src.v4 * ptr[1] + m_dst2src.v5);
int x_rb_src = std::round(m_dst2src.v0 * ptr[2] + m_dst2src.v1 * ptr[3] + m_dst2src.v2);
int y_rb_src = std::round(m_dst2src.v3 * ptr[2] + m_dst2src.v4 * ptr[3] + m_dst2src.v5);
cv::Rect roisrc = cv::Rect(x_lt_src, y_lt_src, x_rb_src - x_lt_src, y_rb_src - y_lt_src) & m_thresh_roisrc;
if (roisrc.width == 0 || roisrc.height == 0)
continue;
// for opencv >=4.7(faster)
// cv::Mat mask_instance;
// cv::resize(cv::Mat(m_mask160, roi160), mask_instance, cv::Size(roisrc.width, roisrc.height), cv::INTER_LINEAR);
// mask_instance = mask_instance > 0.5f;
// cv::cvtColor(mask_instance, mask_instance, cv::COLOR_GRAY2BGR);
// mask_instance.setTo(color, mask_instance);
// cv::addWeighted(mask_instance, 0.45, m_img_canvas(roisrc), 1.0, 0., m_img_canvas(roisrc));
// for opencv >=3.2.0
cv::Mat mask_instance;
cv::resize(cv::Mat(m_mask160, roi160), mask_instance, cv::Size(roisrc.width, roisrc.height), cv::INTER_LINEAR);
mask_instance = mask_instance > 0.5f;
cv::Mat mask_instance_bgr;
cv::cvtColor(mask_instance, mask_instance_bgr, cv::COLOR_GRAY2BGR);
mask_instance_bgr.setTo(color, mask_instance);
cv::addWeighted(mask_instance_bgr, 0.45, m_img_canvas(roisrc), 1.0, 0., m_img_canvas(roisrc));
// label's info
cv::rectangle(imgsBatch[bi], roisrc, color, 2, cv::LINE_AA);
cv::String det_info = m_param.class_names[label] + " " + cv::format("%.4f", ptr[4]);
bbox_points[0][0] = cv::Point(x_lt_src, y_lt_src);
bbox_points[0][1] = cv::Point(x_lt_src + det_info.size() * m_param.char_width, y_lt_src);
bbox_points[0][2] = cv::Point(x_lt_src + det_info.size() * m_param.char_width, y_lt_src - m_param.det_info_render_width);
bbox_points[0][3] = cv::Point(x_lt_src, y_lt_src - m_param.det_info_render_width);
cv::fillPoly(imgsBatch[bi], bbox_point0, num_points, 1, color);
cv::putText(imgsBatch[bi], det_info, bbox_points[0][0], cv::FONT_HERSHEY_DUPLEX, m_param.font_scale, cv::Scalar(255, 255, 255), 1, cv::LINE_AA);
}
}
if (m_param.is_show)
{
cv::imshow(m_param.winname, imgsBatch[bi] + m_img_canvas);
cv::waitKey(cvDelayTime);
}
if (m_param.is_save)
{
cv::imwrite(m_param.save_path + utils::getTimeStamp() + ".jpg", imgsBatch[bi] + m_img_canvas);
}
}
}