CPU-only about interactive-deep-colorization

[junyanz]

I will recommend GPU+CUDA. It can run on a CPU but will be quite slow. You can add --cpu_mode to use your CPU.

[jx2014]

With --cpu_mode, I got to here then it just ended, what does this mean?

I1129 01:00:57.884537 46627 net.cpp:744] Ignoring source layer relu9_2
I1129 01:00:57.884553 46627 net.cpp:744] Ignoring source layer conv9_2norm
I1129 01:00:57.884572 46627 net.cpp:744] Ignoring source layer conv10_1
I1129 01:00:57.884588 46627 net.cpp:744] Ignoring source layer conv1_2_short
I1129 01:00:57.884606 46627 net.cpp:744] Ignoring source layer conv10_1_comb
I1129 01:00:57.884623 46627 net.cpp:744] Ignoring source layer relu10_1_comb
I1129 01:00:57.884639 46627 net.cpp:744] Ignoring source layer conv10_2
I1129 01:00:57.884654 46627 net.cpp:744] Ignoring source layer relu10_2
I1129 01:00:57.884672 46627 net.cpp:744] Ignoring source layer conv10_ab
I1129 01:00:57.884688 46627 net.cpp:744] Ignoring source layer pred_ab_1
I1129 01:00:57.884704 46627 net.cpp:744] Ignoring source layer pred_ab_2
I1129 01:00:57.884783 46627 net.cpp:744] Ignoring source layer loss_ab
I1129 01:00:57.884804 46627 net.cpp:744] Ignoring source layer loss_ab_loss_ab_0_split
I1129 01:00:57.884820 46627 net.cpp:744] Ignoring source layer data_ab_rs
I1129 01:00:57.884835 46627 net.cpp:744] Ignoring source layer gt_ab_313
I1129 01:00:57.925490 46627 net.cpp:744] Ignoring source layer SoftmaxLossPred
I1129 01:00:57.925567 46627 net.cpp:744] Ignoring source layer loss_class_SoftmaxLossPred_0_split
I1129 01:00:57.925587 46627 net.cpp:744] Ignoring source layer loss_log
Setting ab cluster centers in layer: pred_ab
Setting upsampling layer kernel: pred_313_us
./test_imgs/bird_gray.jpg
scale = 2.000000
Killed

[jx2014]

I got it working after increasing ram from 3GB to 4GB in vmware. It runs really slow but it is very cool!!!

[junyanz]

Cool. I am glad that you figured it out. Running on GPUs will be much faster and near real-time.