1# Tencent is pleased to support the open source community by making ncnn available.
2#
3# Copyright (C) 2020 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
15import sys
16import cv2
17import numpy as np
18import ncnn
19from ncnn.model_zoo import get_model
20from ncnn.utils import draw_detection_objects
21
22
23def draw_result(image, class_names, boxes, masks, classes, scores):
24    colors = [
25        [56, 0, 255],
26        [226, 255, 0],
27        [0, 94, 255],
28        [0, 37, 255],
29        [0, 255, 94],
30        [255, 226, 0],
31        [0, 18, 255],
32        [255, 151, 0],
33        [170, 0, 255],
34        [0, 255, 56],
35        [255, 0, 75],
36        [0, 75, 255],
37        [0, 255, 169],
38        [255, 0, 207],
39        [75, 255, 0],
40        [207, 0, 255],
41        [37, 0, 255],
42        [0, 207, 255],
43        [94, 0, 255],
44        [0, 255, 113],
45        [255, 18, 0],
46        [255, 0, 56],
47        [18, 0, 255],
48        [0, 255, 226],
49        [170, 255, 0],
50        [255, 0, 245],
51        [151, 255, 0],
52        [132, 255, 0],
53        [75, 0, 255],
54        [151, 0, 255],
55        [0, 151, 255],
56        [132, 0, 255],
57        [0, 255, 245],
58        [255, 132, 0],
59        [226, 0, 255],
60        [255, 37, 0],
61        [207, 255, 0],
62        [0, 255, 207],
63        [94, 255, 0],
64        [0, 226, 255],
65        [56, 255, 0],
66        [255, 94, 0],
67        [255, 113, 0],
68        [0, 132, 255],
69        [255, 0, 132],
70        [255, 170, 0],
71        [255, 0, 188],
72        [113, 255, 0],
73        [245, 0, 255],
74        [113, 0, 255],
75        [255, 188, 0],
76        [0, 113, 255],
77        [255, 0, 0],
78        [0, 56, 255],
79        [255, 0, 113],
80        [0, 255, 188],
81        [255, 0, 94],
82        [255, 0, 18],
83        [18, 255, 0],
84        [0, 255, 132],
85        [0, 188, 255],
86        [0, 245, 255],
87        [0, 169, 255],
88        [37, 255, 0],
89        [255, 0, 151],
90        [188, 0, 255],
91        [0, 255, 37],
92        [0, 255, 0],
93        [255, 0, 170],
94        [255, 0, 37],
95        [255, 75, 0],
96        [0, 0, 255],
97        [255, 207, 0],
98        [255, 0, 226],
99        [255, 245, 0],
100        [188, 255, 0],
101        [0, 255, 18],
102        [0, 255, 75],
103        [0, 255, 151],
104        [255, 56, 0],
105        [245, 255, 0],
106    ]
107
108    color_index = 0
109
110    for box, mask, label, score in zip(boxes, masks, classes, scores):
111        if score < 0.15:
112            continue
113
114        print(
115            "%s = %.5f at %.2f %.2f %.2f x %.2f\n"
116            % (label, score, box[0], box[1], box[2], box[3])
117        )
118
119        cv2.rectangle(
120            image,
121            (int(box[0]), int(box[1])),
122            (int(box[0] + box[2]), int(int(box[1] + box[3]))),
123            (255, 0, 0),
124        )
125
126        text = "%s %.1f%%" % (class_names[int(label)], score * 100)
127
128        label_size, baseLine = cv2.getTextSize(text, cv2.FONT_HERSHEY_SIMPLEX, 0.5, 1)
129
130        x = box[0]
131        y = box[1] - label_size[1] - baseLine
132        if y < 0:
133            y = 0
134        if x + label_size[0] > image.shape[1]:
135            x = image.shape[1] - label_size[0]
136
137        cv2.rectangle(
138            image,
139            (int(x), int(y)),
140            (int(x + label_size[0]), int(y + label_size[1] + baseLine)),
141            (255, 255, 255),
142            -1,
143        )
144
145        cv2.putText(
146            image,
147            text,
148            (int(x), int(y + label_size[1])),
149            cv2.FONT_HERSHEY_SIMPLEX,
150            0.5,
151            (0, 0, 0),
152        )
153
154        image[mask] = image[mask] * 0.5 + np.array(colors[color_index]) * 0.5
155        color_index += 1
156
157    cv2.imshow("image", image)
158    cv2.waitKey(0)
159
160
161if __name__ == "__main__":
162    if len(sys.argv) != 2:
163        print("Usage: %s [imagepath]" % (sys.argv[0]))
164        sys.exit(0)
165
166    imagepath = sys.argv[1]
167    m = cv2.imread(imagepath)
168    if m is None:
169        print("cv2.imread %s failed\n" % (imagepath))
170        sys.exit(0)
171
172    net = get_model(
173        "yolact",
174        target_size=550,
175        confidence_threshold=0.05,
176        nms_threshold=0.5,
177        keep_top_k=200,
178        num_threads=4,
179        use_gpu=True,
180    )
181
182    boxes, masks, classes, scores = net(m)
183
184    draw_result(m, net.class_names, boxes, masks, classes, scores)
185