DataAugmentation出现的“ValueError: unsupported pickle protocol: 3“问题

如图所示，所描述的错误：

Traceback (most recent call last):
  File "test.py", line 11, in <module>
    bboxes = pkl.load(open("messi_ann.pkl", "rb"))
  File "/usr/lib/python2.7/pickle.py", line 1378, in load
    return Unpickler(file).load()
  File "/usr/lib/python2.7/pickle.py", line 858, in load
    dispatch[key](self)
  File "/usr/lib/python2.7/pickle.py", line 886, in load_proto
    raise ValueError, "unsupported pickle protocol: %d" % proto
ValueError: unsupported pickle protocol: 3

产生的原因：系统自带的“messi_ann.pkl”文件本来就是Python3生成的，我在读取数据的时候用的是python2。

import pickle as pkl

bboxes = pkl.load(open("messi_ann.pkl", "rb"))

解决方法：（1）首先要搞清楚“messi_ann.pkl”文件到底是python2还是python3生成的：

python2生成的要这样导入：  

    import cPickle

    with open('messi_ann.pkl', 'rb') as f:

         samples, labels = cPickle.load(f)

python3生成的要这样导入：

    import pickle

    with open('messi_ann.pkl', 'rb') as f:

        samples, labels = pickle.load(f)

（2）在python3中将数据格式转换成python2可读的数据格式：
   import pickle
with open('./messi_ann.pkl', 'rb') as f:    
   w = pickle.load(f)
pickle.dump(w, open('./messi_annv2.pkl',"wb"), protocol=2)

然后去python2环境中进行操作。

总结：python的版本不同，生成的文件格式不一样。同时要注意import到正确的包文件。

----------------感谢博主：https://blog.csdn.net/c_daofeng/article/details/81129621?utm_source=copy