ncnn

1
// Tencent is pleased to support the open source community by making ncnn available.
2
//
3
// Copyright (C) 2017 THL A29 Limited, a Tencent company. All rights reserved.
4
//
5
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
6
// in compliance with the License. You may obtain a copy of the License at
7
//
8
// https://opensource.org/licenses/BSD-3-Clause
9
//
10
// Unless required by applicable law or agreed to in writing, software distributed
11
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
12
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
13
// specific language governing permissions and limitations under the License.
14

15
#include "net.h"
16

17
#if defined(USE_NCNN_SIMPLEOCV)
18
#include "simpleocv.h"
19
#else
20
#include <opencv2/core/core.hpp>
21
#include <opencv2/highgui/highgui.hpp>
22
#include <opencv2/imgproc/imgproc.hpp>
23
#endif
24
#include <stdio.h>
25
#include <vector>
26

27
struct Object
28
{
29
    cv::Rect_<float> rect;
30
    int label;
31
    float prob;
32
};
33

34
static int detect_mobilenet(const cv::Mat& bgr, std::vector<Object>& objects)
35
{
36
    ncnn::Net mobilenet;
37

38
    mobilenet.opt.use_vulkan_compute = true;
39

40
    // model is converted from https://github.com/chuanqi305/MobileNet-SSD
41
    // and can be downloaded from https://drive.google.com/open?id=0ByaKLD9QaPtucWk0Y0dha1VVY0U
42
    // the ncnn model https://github.com/nihui/ncnn-assets/tree/master/models
43
    if (mobilenet.load_param("mobilenet_ssd_voc_ncnn.param"))
44
        exit(-1);
45
    if (mobilenet.load_model("mobilenet_ssd_voc_ncnn.bin"))
46
        exit(-1);
47

48
    const int target_size = 300;
49

50
    int img_w = bgr.cols;
51
    int img_h = bgr.rows;
52

53
    ncnn::Mat in = ncnn::Mat::from_pixels_resize(bgr.data, ncnn::Mat::PIXEL_BGR, bgr.cols, bgr.rows, target_size, target_size);
54

55
    const float mean_vals[3] = {127.5f, 127.5f, 127.5f};
56
    const float norm_vals[3] = {1.0 / 127.5, 1.0 / 127.5, 1.0 / 127.5};
57
    in.substract_mean_normalize(mean_vals, norm_vals);
58

59
    ncnn::Extractor ex = mobilenet.create_extractor();
60

61
    ex.input("data", in);
62

63
    ncnn::Mat out;
64
    ex.extract("detection_out", out);
65

66
    //     printf("%d %d %d\n", out.w, out.h, out.c);
67
    objects.clear();
68
    for (int i = 0; i < out.h; i++)
69
    {
70
        const float* values = out.row(i);
71

72
        Object object;
73
        object.label = values[0];
74
        object.prob = values[1];
75
        object.rect.x = values[2] * img_w;
76
        object.rect.y = values[3] * img_h;
77
        object.rect.width = values[4] * img_w - object.rect.x;
78
        object.rect.height = values[5] * img_h - object.rect.y;
79

80
        objects.push_back(object);
81
    }
82

83
    return 0;
84
}
85

86
static void draw_objects(const cv::Mat& bgr, const std::vector<Object>& objects)
87
{
88
    static const char* class_names[] = {"background",
89
                                        "aeroplane", "bicycle", "bird", "boat",
90
                                        "bottle", "bus", "car", "cat", "chair",
91
                                        "cow", "diningtable", "dog", "horse",
92
                                        "motorbike", "person", "pottedplant",
93
                                        "sheep", "sofa", "train", "tvmonitor"
94
                                       };
95

96
    cv::Mat image = bgr.clone();
97

98
    for (size_t i = 0; i < objects.size(); i++)
99
    {
100
        const Object& obj = objects[i];
101

102
        fprintf(stderr, "%d = %.5f at %.2f %.2f %.2f x %.2f\n", obj.label, obj.prob,
103
                obj.rect.x, obj.rect.y, obj.rect.width, obj.rect.height);
104

105
        cv::rectangle(image, obj.rect, cv::Scalar(255, 0, 0));
106

107
        char text[256];
108
        sprintf(text, "%s %.1f%%", class_names[obj.label], obj.prob * 100);
109

110
        int baseLine = 0;
111
        cv::Size label_size = cv::getTextSize(text, cv::FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseLine);
112

113
        int x = obj.rect.x;
114
        int y = obj.rect.y - label_size.height - baseLine;
115
        if (y < 0)
116
            y = 0;
117
        if (x + label_size.width > image.cols)
118
            x = image.cols - label_size.width;
119

120
        cv::rectangle(image, cv::Rect(cv::Point(x, y), cv::Size(label_size.width, label_size.height + baseLine)),
121
                      cv::Scalar(255, 255, 255), -1);
122

123
        cv::putText(image, text, cv::Point(x, y + label_size.height),
124
                    cv::FONT_HERSHEY_SIMPLEX, 0.5, cv::Scalar(0, 0, 0));
125
    }
126

127
    cv::imshow("image", image);
128
    cv::waitKey(0);
129
}
130

131
int main(int argc, char** argv)
132
{
133
    if (argc != 2)
134
    {
135
        fprintf(stderr, "Usage: %s [imagepath]\n", argv[0]);
136
        return -1;
137
    }
138

139
    const char* imagepath = argv[1];
140

141
    cv::Mat m = cv::imread(imagepath, 1);
142
    if (m.empty())
143
    {
144
        fprintf(stderr, "cv::imread %s failed\n", imagepath);
145
        return -1;
146
    }
147

148
    std::vector<Object> objects;
149
    detect_mobilenet(m, objects);
150

151
    draw_objects(m, objects);
152

153
    return 0;
154
}
155