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
|
/*
* Copyright (c) 2021 Arm Limited. All rights reserved.
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "UseCaseHandler.hpp"
#include "VisualWakeWordModel.hpp"
#include "Classifier.hpp"
#include "InputFiles.hpp"
#include "UseCaseCommonUtils.hpp"
#include "hal.h"
namespace arm {
namespace app {
/**
* @brief Helper function to load the current image into the input
* tensor.
* @param[in] imIdx Image index (from the pool of images available
* to the application).
* @param[out] inputTensor Pointer to the input tensor to be populated.
* @return true if tensor is loaded, false otherwise.
**/
static bool LoadImageIntoTensor(uint32_t imIdx,
TfLiteTensor *inputTensor);
/* Image inference classification handler. */
bool ClassifyImageHandler(ApplicationContext &ctx, uint32_t imgIndex, bool runAll)
{
auto& platform = ctx.Get<hal_platform &>("platform");
auto& profiler = ctx.Get<Profiler&>("profiler");
constexpr uint32_t dataPsnImgDownscaleFactor = 1;
constexpr uint32_t dataPsnImgStartX = 10;
constexpr uint32_t dataPsnImgStartY = 35;
constexpr uint32_t dataPsnTxtInfStartX = 150;
constexpr uint32_t dataPsnTxtInfStartY = 70;
platform.data_psn->clear(COLOR_BLACK);
time_t infTimeMs = 0;
auto& model = ctx.Get<Model&>("model");
/* If the request has a valid size, set the image index. */
if (imgIndex < NUMBER_OF_FILES) {
if (!SetAppCtxIfmIdx(ctx, imgIndex,"imgIndex")) {
return false;
}
}
if (!model.IsInited()) {
printf_err("Model is not initialised! Terminating processing.\n");
return false;
}
auto curImIdx = ctx.Get<uint32_t>("imgIndex");
TfLiteTensor *outputTensor = model.GetOutputTensor(0);
TfLiteTensor *inputTensor = model.GetInputTensor(0);
if (!inputTensor->dims) {
printf_err("Invalid input tensor dims\n");
return false;
} else if (inputTensor->dims->size < 3) {
printf_err("Input tensor dimension should be >= 3\n");
return false;
}
TfLiteIntArray* inputShape = model.GetInputShape(0);
const uint32_t nCols = inputShape->data[2];
const uint32_t nRows = inputShape->data[1];
const uint32_t nChannels = (inputShape->size == 4) ? inputShape->data[3] : 1;
std::vector<ClassificationResult> results;
do {
/* Strings for presentation/logging. */
std::string str_inf{"Running inference... "};
/* Copy over the data. */
LoadImageIntoTensor(ctx.Get<uint32_t>("imgIndex"), inputTensor);
/* Display this image on the LCD. */
platform.data_psn->present_data_image(
(uint8_t *) inputTensor->data.data,
nCols, nRows, nChannels,
dataPsnImgStartX, dataPsnImgStartY, dataPsnImgDownscaleFactor);
/* If the data is signed. */
if (model.IsDataSigned()) {
image::ConvertImgToInt8(inputTensor->data.data, inputTensor->bytes);
}
/* Display message on the LCD - inference running. */
platform.data_psn->present_data_text(
str_inf.c_str(), str_inf.size(),
dataPsnTxtInfStartX, dataPsnTxtInfStartY, 0);
/* Run inference over this image. */
info("Running inference on image %" PRIu32 " => %s\n", ctx.Get<uint32_t>("imgIndex"),
get_filename(ctx.Get<uint32_t>("imgIndex")));
if (!RunInference(model, profiler)) {
return false;
}
/* Erase. */
str_inf = std::string(str_inf.size(), ' ');
platform.data_psn->present_data_text(
str_inf.c_str(), str_inf.size(),
dataPsnTxtInfStartX, dataPsnTxtInfStartY, 0);
auto& classifier = ctx.Get<Classifier&>("classifier");
classifier.GetClassificationResults(outputTensor, results,
ctx.Get<std::vector <std::string>&>("labels"), 1);
/* Add results to context for access outside handler. */
ctx.Set<std::vector<ClassificationResult>>("results", results);
#if VERIFY_TEST_OUTPUT
arm::app::DumpTensor(outputTensor);
#endif /* VERIFY_TEST_OUTPUT */
if (!image::PresentInferenceResult(platform, results, infTimeMs)) {
return false;
}
profiler.PrintProfilingResult();
IncrementAppCtxIfmIdx(ctx,"imgIndex");
} while (runAll && ctx.Get<uint32_t>("imgIndex") != curImIdx);
return true;
}
static bool LoadImageIntoTensor(const uint32_t imIdx,
TfLiteTensor *inputTensor)
{
const size_t copySz = inputTensor->bytes < IMAGE_DATA_SIZE ?
inputTensor->bytes : IMAGE_DATA_SIZE;
if (imIdx >= NUMBER_OF_FILES) {
printf_err("invalid image index %" PRIu32 " (max: %u)\n", imIdx,
NUMBER_OF_FILES - 1);
return false;
}
const uint32_t nChannels = (inputTensor->dims->size == 4) ? inputTensor->dims->data[3] : 1;
const uint8_t* srcPtr = get_img_array(imIdx);
auto* dstPtr = (uint8_t*)inputTensor->data.data;
if (1 == nChannels) {
/**
* Visual Wake Word model accepts only one channel =>
* Convert image to grayscale here
**/
for (size_t i = 0; i < copySz; ++i, srcPtr += 3) {
*dstPtr++ = 0.2989*(*srcPtr) +
0.587*(*(srcPtr+1)) +
0.114*(*(srcPtr+2));
}
} else {
memcpy(inputTensor->data.data, srcPtr, copySz);
}
debug("Image %" PRIu32 " loaded\n", imIdx);
return true;
}
} /* namespace app */
} /* namespace arm */
|
