I am trying to train a multivariate regression on images.
For classification I used to create a txt file with something like this:
and used convert_imageset tool to create LevelDB or LMDB databases․
Now I have a txt file like this:
image1.jpg 0.1 0.9 0.4
image2.jpg 0.2 0.4 0.2
It seems convert_imageset reads only 2 values from the line. Can I use it to create a corresponding LevelDB / LMDB?
Or maybe should I create 3 LMDBs, one per each variable?! But how I can then combine these 3 variables in a 3-vector in caffe prototxt in order to supply it to the loss layer?