Caffe 中两种读取图片的方式 (python)

读取一张jpg图片（压缩格式）

图表对比

类别	cv2.imread()	caffe.io.load_image()
所在库	import cv2	import caffe
数据类型	0~255 [unsigned char]	0~1 [float]
Channel排列	BGR	RGB
三维数组方向	HWC	HWC

Python Code

    from __future__ import print_function
    import argparse
    import numpy as np
    from func import *
    import caffe
    import struct
    import cv2

    def get_arg():
      parser = argparse.ArgumentParser(description='inference automaticlly')
      parser.add_argument('model', type=str, help='model prototxt for inference' )
      parser.add_argument('weight', type=str, help='model weights(caffemodel) for inference')
      # parser.add_argument('image', type=str, help='image for inference')
      args = parser.parse_args()
      return args

    if __name__ == '__main__':
      args = get_arg()
      # caffe.set_device(0)
      # caffe.set_mode_gpu()
      # net =  caffe.Net(args.model, args.weight, caffe.INT8)
      net =  caffe.Net(args.model, args.weight, caffe.TEST)

      # create transformer for the input called 'data'
      transformer = caffe.io.Transformer({'data':net.blobs['data'].data.shape})

      # load image by caffe
      image = caffe.io.load_image('/home/caffe/autoquan/123_608x608.jpg')

      # HWC -> CHW
      # transformer.set_transpose('data', (2,0,1))
      # transformer.set_mean('data', mu)
      transformer.set_raw_scale('data', 255)
      # RGB -> BGR
      # transformer.set_channel_swap('data', (2,1,0))

      transformed_image = transformer.preprocess('data', image)
      print ("transformed_image = ",transformed_image)

      # load image by opencv2
      image1 = cv2.imread('/home/caffe/autoquan/123_608x608.jpg')
      #  BGR -> RGB
      image1 = cv2.cvtColor(image1, cv2.COLOR_BGR2RGB)

      is_equal = np.all(transformed_image == image1)
      print("cvt 2 RGB = ", image1)

      print (is_equal)