-
Notifications
You must be signed in to change notification settings - Fork 0
/
tflite_infer.c
125 lines (104 loc) · 4.09 KB
/
tflite_infer.c
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
#include <math.h>
#include "config.h"
#include "image.h"
#include "tflite_api.h"
static void error_cb(void* user_data, const char* format, va_list args) {
g_error(format, args);
}
/*
static void tensor_info(const TfLiteTensor* tensor) {
int ndims = TfLiteTensorNumDims(tensor);
g_info("name=%s type=%d dims=%d bytes=%lu", TfLiteTensorName(tensor),
TfLiteTensorType(tensor), ndims, TfLiteTensorByteSize(tensor));
for (int i = 0; i < ndims; ++i) {
g_info("dim %d: %d", i, TfLiteTensorDim(tensor, i));
}
}
*/
enum OutputTensor {
kOutputTensorLocations,
kOutputTensorClasses,
kOutputTensorScores,
kOutputTensorDetections
};
gboolean load_input(const Image* input_img, TfLiteInterpreter* interpreter,
GError** error) {
g_assert_cmpint(1, ==, TfLiteInterpreterGetInputTensorCount(interpreter));
TfLiteTensor* input_tensor = TfLiteInterpreterGetInputTensor(interpreter, 0);
g_autoptr(Image) output_img = image_from_tensor(input_tensor, error);
if (!output_img) return FALSE;
int width, height;
image_size(output_img, &width, &height);
image_resize(input_img, width, height, output_img);
image_swap_rb(output_img, output_img);
return TRUE;
}
void draw_roi(TfLiteInterpreter* interpreter, Image* image) {
g_assert_cmpint(4, ==, TfLiteInterpreterGetOutputTensorCount(interpreter));
const TfLiteTensor* dtn_tensor =
TfLiteInterpreterGetOutputTensor(interpreter, kOutputTensorDetections);
const int num_detections = (int)((float*)TfLiteTensorData(dtn_tensor))[0];
g_info("num detections: %d", num_detections);
const TfLiteTensor* loc_tensor =
TfLiteInterpreterGetOutputTensor(interpreter, kOutputTensorLocations);
const float* locations = (const float*)TfLiteTensorData(loc_tensor);
const TfLiteTensor* cls_tensor =
TfLiteInterpreterGetOutputTensor(interpreter, kOutputTensorClasses);
const float* klasses = (const float*)TfLiteTensorData(cls_tensor);
const TfLiteTensor* scr_tensor =
TfLiteInterpreterGetOutputTensor(interpreter, kOutputTensorScores);
const float* scores = (const float*)TfLiteTensorData(scr_tensor);
int width, height;
image_size(image, &width, &height);
for (int i = 0; i < num_detections; ++i) {
int klass = klasses[i];
int score = round(scores[i] * 100);
int top = round(height * locations[4 * i]);
int left = round(width * locations[4 * i + 1]);
int bottom = round(height * locations[4 * i + 2]);
int right = round(width * locations[4 * i + 3]);
g_info("%d: class=%d score=%d location=[%d %d %d %d]", i, klass, score,
left, top, right - left, bottom - top);
image_draw_roi(image, 1, klass, score, left, top, right - left,
bottom - top);
}
}
int main() {
g_info("tensorflow lite version: %s", TfLiteVersion());
g_autoptr(TfLiteModel) model = TfLiteModelCreateFromFile(TFLITE_MODEL_FILE);
if (!model) {
g_error("TfLiteModelCreateFromFile failed: %s", TFLITE_MODEL_FILE);
return EXIT_FAILURE;
}
g_info("model file: %s", TFLITE_MODEL_FILE);
g_autoptr(TfLiteInterpreterOptions) options =
TfLiteInterpreterOptionsCreate();
TfLiteInterpreterOptionsSetErrorReporter(options, error_cb, NULL);
g_autoptr(TfLiteInterpreter) interpreter =
TfLiteInterpreterCreate(model, options);
if (!interpreter) {
g_error("TfLiteInterpreterCreate failed");
return EXIT_FAILURE;
}
TfLiteStatus status = TfLiteInterpreterAllocateTensors(interpreter);
if (status != kTfLiteOk) {
g_error("TfLiteInterpreterAllocateTensors failed: %d", status);
return EXIT_FAILURE;
}
g_autoptr(GError) error = NULL;
g_autoptr(Image) input_image = image_from_file(INPUT_FILE, &error);
if (!input_image) {
g_error("image_from_file failed: %s", error->message);
}
if (!load_input(input_image, interpreter, &error)) {
g_error("load_input failed: %s", error->message);
}
status = TfLiteInterpreterInvoke(interpreter);
if (status != kTfLiteOk) {
g_error("TfLiteInterpreterInvoke failed: %d", status);
return EXIT_FAILURE;
}
draw_roi(interpreter, input_image);
image_write(input_image, "output.jpeg");
return EXIT_SUCCESS;
}