Created by: NTU-P04922004

Hi,

I have trained a simple mnist model in Caffe and converted it to model.min.json and data.min.bin.

However, when I ran the JNI code provided in the sample code, the result array from the com.baidu.mdl.demo.MDL#predictImage function has a wrong length.

The length of the result array becomes 7840 (10 * 28 * 28) and the correct length should be 10.

Anyone encounter the issue? Any feedback would be great, thank you!

Here are my Caffe model definition and converted json file:

mnist_small.prototxt https://gist.github.com/NTU-P04922004/cfff95e51e4a33ec7c99266613ddadef

model.min.json https://gist.github.com/NTU-P04922004/2da2758871df6abc97a8a98ebbbe9fd0