本文共 1810 字，大约阅读时间需要 6 分钟。

在复制pdf一段文字时,由于格式原因,变成'豆腐块',如下所示.

再进行应用(百度翻译)时比较麻烦.

 The dataset is recorded using a time-of-flight Intel Creative Interactive Gesture Camera and has J = 16 annotated joints. Although the authors pro- vide different artificially rotated training samples, we only use the genuine 22k. The depth images have a high quality with hardly any missing depth val- ues, and sharp outlines with little noise. However, the pose variability is limited compared to the NYU dataset. Also, a relatively large number of samples both from the training and test sets are incorrectly annotated: We evaluated the accuracy and about 36% of the poses from the test set have an annotation error of at least 10 mm.

写了一小段python,对文本进行并行处理

def main():    with open('a.md', 'r+') as obj:        lines = obj.readlines()        strr = ''        for line in lines:            line = line.rstrip()            if len(line)==0:                pass            elif line[-1] == '-':                strr += line[:-1]            else:                strr += line+' '        obj.write(strr)    obj.close()if __name__ == '__main__':    main()

此处'a.md'是ubuntu下随便起的文件名称.(Windows可以改为'a.txt')

注意要把py文件和md文件放在一个文件夹下.

处理后结果:

 The dataset is recorded using a time-of-flight Intel Creative Interactive Gesture Camera and has J = 16 annotated joints. Although the authors provide different artificially rotated training samples, we only use the genuine 22k. The depth images have a high quality with hardly any missing depth values, and sharp outlines with little noise. However, the pose variability is limited compared to the NYU dataset. Also, a relatively large number of samples both from the training and test sets are incorrectly annotated: We evaluated the accuracy and about 36% of the poses from the test set have an annotation error of at least 10 mm. 

转载地址：http://xprbb.baihongyu.com/