Picamera2 库 + TensorFlow Lite 实现实时对象检测

2022年5月16日 树莓派实验室 未分类 1

Raspberry Pi OS Bullseye 版本发布时弃用的 Picamera 以新面貌回归了，那就是 Picamera2 库。

下面介绍使用树莓派摄像头模块，用 Picamera2 库和 TensorFlow Lite 实现实时对象检测。

安装 Picamera2 库

因为目前 Picamera2 库还处于预览版，因此暂时只能通过 GitHub 编译安装。

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
$ sudo apt update
$ sudo apt install -y libboost-dev
$ sudo apt install -y libgnutls28-dev openssl libtiff5-dev
$ sudo apt install -y qtbase5-dev libqt5core5a libqt5gui5 libqt5widgets5
$ sudo apt install -y meson
$ sudo pip3 install pyyaml ply
$ sudo pip3 install --upgrade meson
$ sudo apt install -y libglib2.0-dev libgstreamer-plugins-base1.0-dev
$ git clone --branch picamera2 https://github.com/raspberrypi/libcamera.git
$ cd libcamera
$ meson build --buildtype=release -Dpipelines=raspberrypi -Dipas=raspberrypi -Dv4l2=true -Dgstreamer=enabled -Dtest=false -Dlc-compliance=disabled -Dcam=disabled -Dqcam=enabled -Ddocumentation=disabled -Dpycamera=enabled
$ ninja -C build
$ sudo ninja -C build install
$ cd ~
$ git clone https://github.com/tomba/kmsxx.git
$ cd kmsxx
$ git submodule update --init
$ sudo apt install -y libfmt-dev libdrm-dev
$ meson build
$ ninja -C build
$ cd ~
$ sudo pip3 install pyopengl
$ sudo apt install python3-pyqt5
$ git clone https://git@github.com:raspberrypi/picamera2.git
$ sudo pip3 install opencv-python==4.4.0.46
$ sudo apt install -y libatlas-base-dev
$ sudo pip3 install numpy --upgrade
$ cd ~
$ git clone https://github.com/RaspberryPiFoundation/python-v4l2.git

要跑起来还需要设置 PYTHONPATH 运行环境。比如你需要将下面的内容添加到 .bashrc 文件中。

1
export PYTHONPATH=/home/pi/picamera2:/home/pi/libcamera/build/src/py:/home/pi/kmsxx/build/py:/home/pi/python-v4l2

安装 TensorFlow Lite

由于我们将通过 Python 代码进行推理而不是训练，因此我们可以安装轻量级 TensorFlow Lite 运行时库以及我们需要的其他一些东西：

1
2
3
4
5
6
7
8
$ sudo apt install build-essentials
$ sudo apt install git
$ sudo apt install libatlas-base-dev
$ sudo apt install python3-pip
$ pip3 install tflite-runtime
$ pip3 install opencv-python==4.4.0.46
$ pip3 install pillow
$ pip3 install numpy

开始使用 TensorFlow Lite

安装完所有东西之后，下面我们构建一个演示程序：寻找图像中的苹果和香蕉。

视频播放器

代码会启用摄像头，并将采集到的图像不断传给 TensorFlow 的图像缓冲区。TensorFlow 随之在图像上进行对象检测。如果检测到任何对象，将用矩形框进行标注。

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
import tflite_runtime.interpreter as tflite
 
import sys
import os
import argparse
 
import cv2
import numpy as np
from PIL import Image
from PIL import ImageFont, ImageDraw
 
from qt_gl_preview import *
from picamera2 import *
 
normalSize = (640, 480)
lowresSize = (320, 240)
 
rectangles = []
 
def ReadLabelFile(file_path):
  with open(file_path, 'r') as f:
    lines = f.readlines()
  ret = {}
  for line in lines:
    pair = line.strip().split(maxsplit=1)
    ret[int(pair[0])] = pair[1].strip()
  return ret
 
def DrawRectangles(request):
   stream = request.picam2.stream_map["main"]
   fb = request.request.buffers[stream]
   with fb.mmap(0) as b:
       im = np.array(b, copy=False, dtype=np.uint8).reshape((normalSize[1],normalSize[0], 4))
 
       for rect in rectangles:
          print(rect)
          rect_start = (int(rect[0]*2) - 5, int(rect[1]*2) - 5)
          rect_end = (int(rect[2]*2) + 5, int(rect[3]*2) + 5)
          cv2.rectangle(im, rect_start, rect_end, (0,255,0,0))
          if len(rect) == 5:
            text = rect[4]
            font = cv2.FONT_HERSHEY_SIMPLEX
            cv2.putText(im, text, (int(rect[0]*2) + 10, int(rect[1]*2) + 10), font, 1, (255,255,255),2,cv2.LINE_AA)
       del im
 
def InferenceTensorFlow( image, model, output, label=None):
   global rectangles
 
   if label:
       labels = ReadLabelFile(label)
   else:
       labels = None
 
   interpreter = tflite.Interpreter(model_path=model, num_threads=4)
   interpreter.allocate_tensors()
 
   input_details = interpreter.get_input_details()
   output_details = interpreter.get_output_details()
   height = input_details[0]['shape'][1]
   width = input_details[0]['shape'][2]
   floating_model = False
   if input_details[0]['dtype'] == np.float32:
       floating_model = True
 
   rgb = cv2.cvtColor(image,cv2.COLOR_GRAY2RGB)
   initial_h, initial_w, channels = rgb.shape
 
   picture = cv2.resize(rgb, (width, height))
 
   input_data = np.expand_dims(picture, axis=0)
   if floating_model:
      input_data = (np.float32(input_data) - 127.5) / 127.5
 
   interpreter.set_tensor(input_details[0]['index'], input_data)
 
   interpreter.invoke()
 
   detected_boxes = interpreter.get_tensor(output_details[0]['index'])
   detected_classes = interpreter.get_tensor(output_details[1]['index'])
   detected_scores = interpreter.get_tensor(output_details[2]['index'])
   num_boxes = interpreter.get_tensor(output_details[3]['index'])
 
   rectangles = []
   for i in range(int(num_boxes)):
      top, left, bottom, right = detected_boxes[0][i]
      classId = int(detected_classes[0][i])
      score = detected_scores[0][i]
      if score > 0.5:
          xmin = left * initial_w
          ymin = bottom * initial_h
          xmax = right * initial_w
          ymax = top * initial_h
          box = [xmin, ymin, xmax, ymax]
          rectangles.append(box)
          if labels:
              print(labels[classId], 'score = ', score)
              rectangles[-1].append(labels[classId])
          else:
              print ('score = ', score)
 
def main():
    parser = argparse.ArgumentParser()
    parser.add_argument('--model', help='Path of the detection model.', required=True)
    parser.add_argument('--label', help='Path of the labels file.')
    parser.add_argument('--output', help='File path of the output image.')
    args = parser.parse_args()
 
    if ( args.output):
      output_file = args.output
    else:
      output_file = 'out.jpg'
 
    if ( args.label ):
      label_file = args.label
    else:
      label_file = None
 
    picam2 = Picamera2()
    preview = QtGlPreview(picam2)
    config = picam2.preview_configuration(main={"size": normalSize},
                                          lores={"size": lowresSize, "format": "YUV420"})
    picam2.configure(config)
 
    stride = picam2.stream_configuration("lores")["stride"]
    picam2.request_callback = DrawRectangles
 
    picam2.start()
 
    while True:
        buffer = picam2.capture_buffer("lores")
        grey = buffer[:stride*lowresSize[1]].reshape((lowresSize[1], stride))
        result = InferenceTensorFlow( grey, args.model, output_file, label_file )
 
if __name__ == '__main__':
  main()

via

12,903


1 评论

向草芥进行回复 取消回复

你的邮件地址不会公开


*