cnn_cls.cpp
5.51 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
#include "cnn_cls.h"
#include <iostream>
#include "acl/acl.h"
#include "model_process.h"
#include "sy_errorinfo.h"
#include <time.h>
#include <sys/time.h>
#include <algorithm>
using namespace std;
namespace atlas_utils {
int CnnCls::Init(const char* modelPath) {
ACL_CALL(aclrtGetRunMode(&runMode_), SY_SUCCESS, SY_FAILED);//获取当前昇腾AI软件栈的运行模式，根据不同的运行模式，后续的接口调用方式不同
ACL_CALL(model_.LoadModelFromFileWithMem(modelPath), SY_SUCCESS, SY_FAILED);//从文件加载离线模型数据，由用户自行管理模型运行的内存
ACL_CALL(model_.CreateDesc(), SY_SUCCESS, SY_FAILED);//获取模型的描述信息
ACL_CALL(model_.CreateOutput(outDims_), SY_SUCCESS, SY_FAILED);
ACL_CALL(model_.GetInputDims(inDims_), SY_SUCCESS, SY_FAILED);
modelHeight_ = inDims_[0][1];
modelWidth_ = inDims_[0][2];
return SY_SUCCESS;
}
// double msecond1() {
// struct timeval tv;
// gettimeofday(&tv, 0);
// return (tv.tv_sec * 1000.0 + tv.tv_usec / 1000.0);
// }
int CnnCls::Inference(ImageData& input) {
model_.CreateInput(input.data.get(), input.size);
//double t1, t2;
//t1 = msecond1();
ACL_CALL(model_.Execute(), SY_SUCCESS, SY_FAILED);
//t2 = msecond1();
//printf("debug forward time: %.2f\n", t2 - t1);
model_.DestroyInput(); //需调用CreateInput的销毁类接口DestroyInput！！！
return SY_SUCCESS;
}
int CnnCls::GetInputWidth() {
return modelWidth_;
}
int CnnCls::GetInputHeight() {
return modelHeight_;
}
int CnnCls::PostProcess(vector<float>& results) {
aclmdlDataset* modelOutput = model_.GetModelOutputData();
int outDatasetNum = aclmdlGetDatasetNumBuffers(modelOutput);
for (int i = 0; i < outDatasetNum; i++) {
aclDataBuffer* dataBuffer = aclmdlGetDatasetBuffer(modelOutput, i);
if (dataBuffer == nullptr) {
return SY_FAILED;
}
uint32_t dataBufferSize = aclGetDataBufferSize(dataBuffer);
void* data = aclGetDataBufferAddr(dataBuffer);
if (data == nullptr) {
return SY_FAILED;
}
int length = dataBufferSize/sizeof(float);
float outInfo[length];
if (runMode_ == ACL_HOST) {
ACL_CALL(aclrtMemcpy(outInfo, sizeof(outInfo), data, sizeof(outInfo), ACL_MEMCPY_DEVICE_TO_HOST),
ACL_SUCCESS, SY_FAILED);
} else {
ACL_CALL(aclrtMemcpy(outInfo, sizeof(outInfo), data, sizeof(outInfo), ACL_MEMCPY_DEVICE_TO_DEVICE),ACL_SUCCESS, SY_FAILED);
//return SY_FAILED;
}
int argmax = std::distance(outInfo, std::max_element(outInfo, outInfo + length));
if(outInfo[argmax] < config.confThr){
// printf("outInfo[argmax]:%f\n",outInfo[argmax]);
// printf("config.confThr:%f\n",config.confThr);
results.emplace_back(-1);
results.emplace_back(0);
// INFO_LOG("vColor is low confidence!");
}
else{
results.emplace_back(argmax);
results.emplace_back(outInfo[argmax]);
}
/*
for(uint32_t b = 0; b < length; b++) {
results.emplace_back(outInfo[b]);
}*/
}
return SY_SUCCESS;
}
int CnnCls::PostProcess_batch(vector<vector<float>>& results) {
aclmdlDataset* modelOutput = model_.GetModelOutputData();
int outDatasetNum = aclmdlGetDatasetNumBuffers(modelOutput);
const int batchsize = outDims_[0][0];
// 结果拷贝
vector<vector<float>> all_res;
for (int i = 0; i < outDatasetNum; i++) {
aclDataBuffer* dataBuffer = aclmdlGetDatasetBuffer(modelOutput, i);
if (dataBuffer == nullptr) {
return SY_FAILED;
}
uint32_t dataBufferSize = aclGetDataBufferSize(dataBuffer);
void* data = aclGetDataBufferAddr(dataBuffer);
if (data == nullptr) {
return SY_FAILED;
}
int length = dataBufferSize/sizeof(float);
float outInfo[length];
if (runMode_ == ACL_HOST) {
ACL_CALL(aclrtMemcpy(outInfo, sizeof(outInfo), data, sizeof(outInfo), ACL_MEMCPY_DEVICE_TO_HOST),ACL_SUCCESS, SY_FAILED);
} else {
ACL_CALL(aclrtMemcpy(outInfo, sizeof(outInfo), data, sizeof(outInfo), ACL_MEMCPY_DEVICE_TO_DEVICE),ACL_SUCCESS, SY_FAILED);
}
vector<float> res;
res.assign(outInfo, outInfo + length);
all_res.emplace_back(res);
}
// 按[b][cls]的形式返回
for (int b = 0; b < batchsize; b ++) {
vector<float> result;
for (const auto& outInfo : all_res) {
int single_length = outInfo.size() / batchsize;
// printf("batchsize:%d,single_length:%d\n",batchsize, single_length);
int argmax = std::distance(outInfo.begin()+b*single_length, std::max_element(outInfo.begin()+b*single_length, outInfo.begin() + (b+1)*single_length));
float score = outInfo[b*single_length+argmax];
if(score < config.confThr) {
// printf("score:%f, confThr\n",score, config.confThr);
result.emplace_back(-1);
result.emplace_back(0);
}
else {
result.emplace_back(argmax);
result.emplace_back(score);
}
}
results.emplace_back(result);
}
return SY_SUCCESS;
}
void CnnCls::Release() {
model_.Unload();
model_.DestroyDesc();
model_.DestroyOutput();
}
}