JeVoisBase  1.18
JeVois Smart Embedded Machine Vision Toolkit Base Modules
Share this page:
demo.py
Go to the documentation of this file.
1 # This file is part of OpenCV Zoo project.
2 # It is subject to the license terms in the LICENSE file found in the same directory.
3 #
4 # Copyright (C) 2021, Shenzhen Institute of Artificial Intelligence and Robotics for Society, all rights reserved.
5 # Third party copyrights are property of their respective owners.
6 
7 import argparse
8 
9 import numpy as np
10 import cv2 as cv
11 
12 from yunet import YuNet
13 
14 def str2bool(v):
15  if v.lower() in ['on', 'yes', 'true', 'y', 't']:
16  return True
17  elif v.lower() in ['off', 'no', 'false', 'n', 'f']:
18  return False
19  else:
20  raise NotImplementedError
21 
22 backends = [cv.dnn.DNN_BACKEND_OPENCV, cv.dnn.DNN_BACKEND_CUDA]
23 targets = [cv.dnn.DNN_TARGET_CPU, cv.dnn.DNN_TARGET_CUDA, cv.dnn.DNN_TARGET_CUDA_FP16]
24 help_msg_backends = "Choose one of the computation backends: {:d}: OpenCV implementation (default); {:d}: CUDA"
25 help_msg_targets = "Chose one of the target computation devices: {:d}: CPU (default); {:d}: CUDA; {:d}: CUDA fp16"
26 try:
27  backends += [cv.dnn.DNN_BACKEND_TIMVX]
28  targets += [cv.dnn.DNN_TARGET_NPU]
29  help_msg_backends += "; {:d}: TIMVX"
30  help_msg_targets += "; {:d}: NPU"
31 except:
32  print('This version of OpenCV does not support TIM-VX and NPU. Visit https://gist.github.com/fengyuentau/5a7a5ba36328f2b763aea026c43fa45f for more information.')
33 
34 parser = argparse.ArgumentParser(description='YuNet: A Fast and Accurate CNN-based Face Detector (https://github.com/ShiqiYu/libfacedetection).')
35 parser.add_argument('--input', '-i', type=str, help='Path to the input image. Omit for using default camera.')
36 parser.add_argument('--model', '-m', type=str, default='face_detection_yunet_2022mar.onnx', help='Path to the model.')
37 parser.add_argument('--backend', '-b', type=int, default=backends[0], help=help_msg_backends.format(*backends))
38 parser.add_argument('--target', '-t', type=int, default=targets[0], help=help_msg_targets.format(*targets))
39 parser.add_argument('--conf_threshold', type=float, default=0.9, help='Filter out faces of confidence < conf_threshold.')
40 parser.add_argument('--nms_threshold', type=float, default=0.3, help='Suppress bounding boxes of iou >= nms_threshold.')
41 parser.add_argument('--top_k', type=int, default=5000, help='Keep top_k bounding boxes before NMS.')
42 parser.add_argument('--save', '-s', type=str, default=False, help='Set true to save results. This flag is invalid when using camera.')
43 parser.add_argument('--vis', '-v', type=str2bool, default=True, help='Set true to open a window for result visualization. This flag is invalid when using camera.')
44 args = parser.parse_args()
45 
46 def visualize(image, results, box_color=(0, 255, 0), text_color=(0, 0, 255), fps=None):
47  output = image.copy()
48  landmark_color = [
49  (255, 0, 0), # right eye
50  ( 0, 0, 255), # left eye
51  ( 0, 255, 0), # nose tip
52  (255, 0, 255), # right mouth corner
53  ( 0, 255, 255) # left mouth corner
54  ]
55 
56  if fps is not None:
57  cv.putText(output, 'FPS: {:.2f}'.format(fps), (0, 15), cv.FONT_HERSHEY_SIMPLEX, 0.5, text_color)
58 
59  for det in (results if results is not None else []):
60  bbox = det[0:4].astype(np.int32)
61  cv.rectangle(output, (bbox[0], bbox[1]), (bbox[0]+bbox[2], bbox[1]+bbox[3]), box_color, 2)
62 
63  conf = det[-1]
64  cv.putText(output, '{:.4f}'.format(conf), (bbox[0], bbox[1]+12), cv.FONT_HERSHEY_DUPLEX, 0.5, text_color)
65 
66  landmarks = det[4:14].astype(np.int32).reshape((5,2))
67  for idx, landmark in enumerate(landmarks):
68  cv.circle(output, landmark, 2, landmark_color[idx], 2)
69 
70  return output
71 
72 if __name__ == '__main__':
73  # Instantiate YuNet
74  model = YuNet(modelPath=args.model,
75  inputSize=[320, 320],
76  confThreshold=args.conf_threshold,
77  nmsThreshold=args.nms_threshold,
78  topK=args.top_k,
79  backendId=args.backend,
80  targetId=args.target)
81 
82  # If input is an image
83  if args.input is not None:
84  image = cv.imread(args.input)
85  h, w, _ = image.shape
86 
87  # Inference
88  model.setInputSize([w, h])
89  results = model.infer(image)
90 
91  # Print results
92  print('{} faces detected.'.format(results.shape[0]))
93  for idx, det in enumerate(results):
94  print('{}: {:.0f} {:.0f} {:.0f} {:.0f} {:.0f} {:.0f} {:.0f} {:.0f} {:.0f} {:.0f} {:.0f} {:.0f} {:.0f} {:.0f}'.format(
95  idx, *det[:-1])
96  )
97 
98  # Draw results on the input image
99  image = visualize(image, results)
100 
101  # Save results if save is true
102  if args.save:
103  print('Resutls saved to result.jpg\n')
104  cv.imwrite('result.jpg', image)
105 
106  # Visualize results in a new window
107  if args.vis:
108  cv.namedWindow(args.input, cv.WINDOW_AUTOSIZE)
109  cv.imshow(args.input, image)
110  cv.waitKey(0)
111  else: # Omit input to call default camera
112  deviceId = 0
113  cap = cv.VideoCapture(deviceId)
114  w = int(cap.get(cv.CAP_PROP_FRAME_WIDTH))
115  h = int(cap.get(cv.CAP_PROP_FRAME_HEIGHT))
116  model.setInputSize([w, h])
117 
118  tm = cv.TickMeter()
119  while cv.waitKey(1) < 0:
120  hasFrame, frame = cap.read()
121  if not hasFrame:
122  print('No frames grabbed!')
123  break
124 
125  # Inference
126  tm.start()
127  results = model.infer(frame) # results is a tuple
128  tm.stop()
129 
130  # Draw results on the input image
131  frame = visualize(frame, results, fps=tm.getFPS())
132 
133  # Visualize results in a new Window
134  cv.imshow('YuNet Demo', frame)
135 
136  tm.reset()
137 
demo.str2bool
str2bool
Definition: demo.py:43
demo.int
int
Definition: demo.py:37
demo.visualize
def visualize(image, results, box_color=(0, 255, 0), text_color=(0, 0, 255), fps=None)
Definition: demo.py:46
yunet.YuNet
Definition: yunet.py:12