mnist数据集Lenet培训

Question

我是初出茅庐的。为了训练Lenet使用mnist数据集，我将数据下载到caffe中的data/mnist。当试图使用./examples/mnist/create_mnist.sh转换为lmdb时，我得到的错误如下：

 Creating lmdb...
F0225 16:39:23.937786 23330 convert_mnist_data.cpp:45] Check failed: image_file Unable to open file data/mnist/train-images-idx3-ubyte
*** Check failure stack trace: ***
    @     0x7f5d0f0e8daa  (unknown)
    @     0x7f5d0f0e8ce4  (unknown)
    @     0x7f5d0f0e86e6  (unknown)
    @     0x7f5d0f0eb687  (unknown)
    @           0x4045df  convert_dataset()
    @           0x402943  main
    @     0x7f5d0de99ec5  (unknown)
    @           0x40298c  (unknown)
    @              (nil)  (unknown)
Aborted (core dumped)
F0225 16:39:24.025714 23334 convert_mnist_data.cpp:45] Check failed: image_file Unable to open file data/mnist/t10k-images-idx3-ubyte
*** Check failure stack trace: ***
    @     0x7faa1af57daa  (unknown)
    @     0x7faa1af57ce4  (unknown)
    @     0x7faa1af576e6  (unknown)
    @     0x7faa1af5a687  (unknown)
    @           0x4045df  convert_dataset()
    @           0x402943  main
    @     0x7faa19d08ec5  (unknown)
    @           0x40298c  (unknown)
    @              (nil)  (unknown)
Aborted (core dumped)
Done.

有人能帮我吗？究竟是什么问题？

Answer 1

尽管脚本由于未知的原因而失败，但手动编译仍然有效。

Answer 2

您可以使用绝对路径，而不是像data/mnist/t10k-images-idx3-ubyte，.caffe有时无法得到它。

Answer 3

重命名您的mnist数据集，因为您下载的mnist数据集可能是"t10k-images.idx3-ubyte“，但它是create_mnist.sh中的”t10k-映像-idx3-ubyte“，所以您必须使用‘-’而不是'.‘。