hi!
I am running a network on input type "ImageData" , and I am getting labels that are not what I wrote.
test text file looks right:
<path>_c179_id262942_g7from10_f1785_mirror.png 0
<path>_c109_id109102_g2from2_f2496_mirror.png 0
<path>_c109_id109104_g4from6_f2304_mirror.png 0
<path>_c55_id56119_g3from7_f2634_mirror.png 1
<path>_c305_id11495_g4from6_f570_mirror.png 1
<path>_c233_id1902_g1from1_f327_mirror.png 0
<path>_c109_id108946_g4from5_f936_mirror.png 1
<path>_c233_id1896_g4from6_f147.png 0
<path>_c233_id1923_g1from1_f459_mirror.png 0
<path>_c233_id1947_g4from5_f591.png 1
test data layer looks like this:
layer {
name: "data_2"
#type: "Data"
type: "ImageData"
top: "data_2"
top: "dummy_sim_labels"
include {
phase: TEST
}
transform_param {
mean_file: "<path>_mean.binaryproto"
scale: 0.00393
}
image_data_param {
source: "<path>_train2_b.txt"
batch_size: 100
}
}
test label print to screen looks like this:
I0725 11:16:22.082417 19373 solver.cpp:404] Test net output #0: dummy_sim_labels = 0.5
I0725 11:16:22.767827 19373 solver.cpp:404] Test net output #1: dummy_sim_labels = 0.46
I0725 11:16:22.767844 19373 solver.cpp:404] Test net output #2: dummy_sim_labels = 0.5
I0725 11:16:22.767848 19373 solver.cpp:404] Test net output #3: dummy_sim_labels = 0.44
I0725 11:16:22.767849 19373 solver.cpp:404] Test net output #4: dummy_sim_labels = 0.48
I0725 11:16:22.767851 19373 solver.cpp:404] Test net output #5: dummy_sim_labels = 0.58
the test text and train text were creates with the same script, different image lists, from the same image data set, and are in the same location.
train text file looks right:
<path>_c120_id173432_g2from7_f1275_mirror.png 0
<path>_c110_id117859_g1from2_f3374_mirror.png 0
<path>_c136_id204229_g5from10_f1435.png 1
<path>_c158_id231821_g6from6_f1456_mirror.png 1
<path>_c110_id117767_g5from17_f1435_mirror.png 1
<path>_c108_id102891_g5from5_f2304.png 0
<path>_c425_id24487_g4from4_f90_mirror.png 1
<path>_c539_id42073_g2from2_f296_mirror.png 1
<path>_c135_id201877_g1from1_f2928_mirror.png 0
<path>_c42_id32617_g4from8_f4566_mirror.png 0
train data layer looks like this:
layer {
name: "data_2"
#type: "Data"
type: "ImageData"
top: "data_2"
top: "dummy_sim_labels"
include {
phase: TRAIN
}
transform_param {
mean_file: "<path>_mean.binaryproto"
scale: 0.00393
}
image_data_param {
source: "<path>_train1_b.txt"
batch_size: 100
}
}
train label print to screen looks like this:
I0725 11:10:47.576225 19373 solver.cpp:244] Train net output #48: dummy_sim_labels = 1
I0725 11:10:47.576226 19373 solver.cpp:244] Train net output #49: dummy_sim_labels = 0
I0725 11:10:47.576228 19373 solver.cpp:244] Train net output #50: dummy_sim_labels = 1
I0725 11:10:47.576231 19373 solver.cpp:244] Train net output #51: dummy_sim_labels = 0
I0725 11:10:47.576232 19373 solver.cpp:244] Train net output #52: dummy_sim_labels = 1
I0725 11:10:47.576234 19373 solver.cpp:244] Train net output #53: dummy_sim_labels = 1
* <path> in reality is some text that I replaced to hide names just for this post...