I am doing regression using a fine-tuned net, on the lines of caffe flickr-style example, I have changed the num outputs in the last layer to 1, but when testing it on an image using the matlab wrapper function matcaffe_demo()
. It returns 10 outputs corresponding to a single image while it should return only 1.
This is my final layer in 'deploy.prototxt'
layer {
name: "my-fc8"
type: "InnerProduct"
bottom: "fc7"
top: "my-fc8"
# lr_mult is set to higher than for other layers, because this layer is starting from random while the others are already trained
param {
lr_mult: 10
decay_mult: 1
}
param {
lr_mult: 20
decay_mult: 0
}
inner_product_param {
num_output: 1
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
value: 0
}
}
As one can see, the num_outputs
is 1, the problem is when I test the finetuned net on an image using matcaffe_demo()
, it gives 10 output labels instead of 1. Can anyone help me understand what is happening here. Thanks in advance.