datum_height == data_mean_.height() when running solver.prototxt?

[archan...@gmail.com]

Hey,

I'm relatively new to caffe and am trying to run solver.prototxt on my own data (./build/tools/caffe train --solver=path/to/solver.prototxt ) and am currently getting this issue:

data_transformer.cpp:63] Check failed: datum_height == data_mean_.height() (760 vs. 256)

So the dimensions of something are wrong but I'm not sure how to fix it. It's probably after the layer conv1.

Here's my train_val.prototxt and solver.prototxt--slightly modified from what's here. 
My images are 256x256 -- some things are admittedly pretty weird in the files because my dataset is stupid levels of small at this point (I'm mainly using it to familiarize myself with using caffe for my own datasets).
solver.txt
train_val.txt

Any help would be great!

[archan...@gmail.com]

resolved. 

[karim sayadi]

Hi, 

Can you please explain how you resolved that ? 

Thank you in advance. 

[dejian...@gmail.com]

The problem is told that your input image size is not equal your mean file. As for how to solve it, it depends on what type of input data you gave to caffe, what you have to do is change you input data size equal to your mean file size.