Same matching accuracy always

[skazmi1]

I created my idx3,idx1 files in mnisten. I then trained them in tiny dnn using example_mnist_train and tested them with example_mnist_test. I always get the same accuracy regardless of what testing image I use. Has anyone else encountered this before?

P.S I am using black on white backgrounds like it is supposed to be.

[lenghanyan]

请问一下，这个代码你是在window还是linux上面运行的？？为什么我在win上面运行，在cmd上输入参数出现闪退现象？能否给我一点帮助？谢谢