NanoPI-T3上搭建人工智能环境

相关资源汇总：

TensorFlow Lite 官网

TensorFlow Lite 官方指南

一、摄像头数据读取

直接采用OPENCV读取摄像头数据，参考NanoPI T3官方文档，由于这里我是先装的TensorFlowLite，其中需要升级numpy，这里遇到了一点问题，提前说明下：

该板子python中的包通过pip进行安装或卸载后，需要手动卸载掉当前的包才能升级或卸载：

#升级到最新
pip install packname
#卸载
pip uninstall packname
#查看当前包
pip list

• 有时候可能需要利用sudo，或者是sudo -H
• pip的，默认源是官方源，可以使用-i https://pypi.tuna.tsinghua.edu.cn/simple some-package命令来临时采用清华源（由于NanoPi T3中ubuntu16.04所使用的python最新版本较低，所以这里不推荐升级pip）

这里给出一个调用摄像头的demo：

import numpy as np
import cv2 as cv
cap = cv.VideoCapture(0)
if not cap.isOpened():
    print("Cannot open camera")
    exit()

cap.set(cv.CAP_PROP_FRAME_HEIGHT,240)
cap.set(cv.CAP_PROP_FRAME_WIDTH,320)

'''使用cv.VideoCapture.set设置摄像头分辨率是出现错误'''
while True:
    # Capture frame-by-frame
    ret, frame = cap.read()
    # if frame is read correctly ret is True
    if not ret:
        print("Can't receive frame (stream end?). Exiting ...")
        break
    cv.imshow('frame', frame)
    if cv.waitKey(1) == ord('q'):
        break
# When everything done, release the capture
cap.release()
cv.destroyAllWindows()

• 注意：调用set函数时，如果摄像头不支持此类参数会报类似于以下的错误：

  [ WARN:0] global /root/opencv/modules/videoio/src/cap_gstreamer.cpp (933) open OpenCV | GStreamer warning: Cannot query video position: status=0, value=-1, duration=-1
  [ WARN:0] global /root/opencv/modules/videoio/src/cap_gstreamer.cpp (1757) handleMessage OpenCV | GStreamer warning: Embedded video playback halted; module v4l2src0 reported: Internal data flow error.
  [ WARN:0] global /root/opencv/modules/videoio/src/cap_gstreamer.cpp (515) startPipeline OpenCV | GStreamer warning: unable to start pipeline
  [ WARN:0] global /root/opencv/modules/videoio/src/cap_gstreamer.cpp (1055) setProperty OpenCV | GStreamer warning: no pipeline
  Can't receive frame (stream end?). Exiting ...
  [ WARN:0] global /root/opencv/modules/videoio/src/cap_gstreamer.cpp (480) isPipelinePlaying OpenCV | GStreamer warning: GStreamer: pipeline have not been created

二、TensorFlowLite安装

参考官方，这里采用python来操作，官方给出了python入门，这里可以直接参考：

如果需要C/C++操作，可以参考在ARM64上构建TensorFlowLite，需要自己通过交叉编译工具链进行编译

通过下面的命令确定python版本和机器位数：

#python版本
python3 --version
#机器位数
uname -a

对于NanoPi T3来说是ARM64，python3.5，故采用以下命令直接安装：

sudo pip3 install https://dl.google.com/coral/python/tflite_runtime-2.1.0.post1-cp35-cp35m-linux_aarch64.whl

• 期间可以能需要numpy版本更新，详见上文
• 安装的是tflite_runtime，仅用于解释TensorFlow中的网络
• 可以通过转化器将其他TensorFlow模型转换为TensorFlow Lite模型
• 关于Interpreter的更多API可以参考在Python中加载并运行模型

安装好后可以通过官方针对树莓派的demo来进行：

• 图像分类
• 对象检测

但是树莓派上采用picamera调用摄像头，需要将上述代码中的picamera换为opencv部分，下述代码为修改后可行的图像分类demo：

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import argparse
import io
import time
import numpy as np
import cv2 as cv

from PIL import Image
from tflite_runtime.interpreter import Interpreter

cap = cv.VideoCapture(0)
if not cap.isOpened():
    print("Cannot open camera")
    exit()

cap.set(cv.CAP_PROP_FRAME_HEIGHT,480)
cap.set(cv.CAP_PROP_FRAME_WIDTH,640)

cv.namedWindow("Video")
font = cv.FONT_HERSHEY_SIMPLEX
color = (0, 255, 0)#BGR

def load_labels(path):
    with open(path, 'r') as f:
        return {i: line.strip() for i, line in enumerate(f.readlines())}

def set_input_tensor(interpreter, image):
    tensor_index = interpreter.get_input_details()[0]['index']
    input_tensor = interpreter.tensor(tensor_index)()[0]
    input_tensor[:, :] = image

def classify_image(interpreter, image, top_k=1):
    """Returns a sorted array of classification results."""
    set_input_tensor(interpreter, image)
    interpreter.invoke()
    output_details = interpreter.get_output_details()[0]
    output = np.squeeze(interpreter.get_tensor(output_details['index']))

    # If the model is quantized (uint8 data), then dequantize the results
    if output_details['dtype'] == np.uint8:
        scale, zero_point = output_details['quantization']
        output = scale * (output - zero_point)

    ordered = np.argpartition(-output, top_k)
    return [(i, output[i]) for i in ordered[:top_k]]

def main():
    parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
    parser.add_argument('--model', help='File path of .tflite file.', required=True)
    parser.add_argument('--labels', help='File path of labels file.', required=True)
    args = parser.parse_args()

    labels = load_labels(args.labels)

    interpreter = Interpreter(args.model)
    interpreter.allocate_tensors()
    _, height, width, _ = interpreter.get_input_details()[0]['shape']

    while True:
        # Capture frame-by-frame
        ret, frame = cap.read()
        # if frame is read correctly ret is True
        if not ret:
            print("Can't receive frame (stream end?). Exiting ...")
            break

        image = Image.fromarray(cv.cvtColor(frame,cv.COLOR_BGR2RGB))
        image = image.resize((width, height),Image.ANTIALIAS)

        start_time = time.time()
        results = classify_image(interpreter, image)
        elapsed_ms = (time.time() - start_time) * 1000
        label_id, prob = results[0]

        image = cv.cvtColor(np.array(image),cv.COLOR_RGB2BGR)
        cv.putText(image, 'labels=%s'%(labels[label_id]), (0, 40), font, 0.8, color)
        if prob >= 0.5:
            cv.putText(image, 'prob=%.2f'%(prob), (0, 80), font, 0.8, (0, 0, 255)) #BGR
        else:
            cv.putText(image, 'prob=%.2f'%(prob), (0, 80), font, 0.8, color)
        cv.putText(image, 'ms=%.1f'%(elapsed_ms), (0, 120), font, 0.8, color)
        cv.imshow('frame', image)
        if cv.waitKey(1) == ord('q'):
            cap.release()
            break

if __name__ == '__main__':
    main()
    cv.destroyAllWindows()