caffe - Input data, train and test set

Asked Feb 16 '16 at 00:04

Active Feb 16 '16 at 00:04

Viewed 37 times

RAW data dim :

1) 64 x 45 x 18432 in .mat format... ---> Class1

1 x 18432 in .mat format... ---> labels

2) 64 x 45 x 18432 in .mat format... ---> Class2

1 x 18432 in .mat format... ---> labels

3) 64 x 45 x 18432 in .mat format... ---> Class3

1 x 18432 in .mat format... ---> labels

How to use/convert it for caffe as a input... using python?

asked Feb 16 '16 at 00:04

shivang patel

1

`64`-by-`45` is a gray scale image (single channel), for RGB image you need `64`-by-`45`-by-`3`... – Shai Feb 16 '16 at 10:06
1

feeding caffe data from matlab through hdf5 should not make any difference. You need to feed the data in dimensions `Batch`-by-`channel`-by-`height`-by-`width`. – Shai Feb 16 '16 at 10:07
Ok.... That's very helpful..... – shivang patel Feb 16 '16 at 10:08
Right now, I am train network with with 2 class data... but accuracy is constant 1 after first iteration ! Why is that happened ? – shivang patel Feb 17 '16 at 12:01
this seems to be a new question. – Shai Feb 17 '16 at 12:02
Ok... I make it new one... – shivang patel Feb 17 '16 at 12:04

0 Answers0