xcyan / eccv16_attr2img Goto Github PK

How to get this mask images of lfw for python and tensorflow.

Hi Xinchen,

I just got to your wonderful paper Attribute2Image and I have a question about it. It will be great if you can answer it.

I saw that disCVAE decomposes the image generation into a foreground image generation and a background generation. The final image is the composition of the two. My question is do we need ground truth segmentation masks of training images to implement disCVAE?

It seem your code use the ground truth segmentation to train disCVAE?

Thanks in advance!

I was finally able to train the networks using leased time in a GPU cluster, saved the trained networks (the net-epoch-100.t7 files), and run the decoder for testing. Everything runs fine in the GPU cluster, but when trying to use the trained networks on a machine without GPU the net-epoch-100.t7 files cannot be loaded. Speed is indeed needed for training, but, after the network has been trained, just feeding attributes + latent variables to the decoder to get synthetic images is a different matter. Is there any way to convert the net-epoch-100.t7 files into something equivalent that does not require using a GPU? Thank you!

i can not download dataset by https://umich.box.com/shared/static/dqxmrsy8xirkw3tjqcql25900wz7lvjj.zip
could you give me another dataset download link or lfw dataset pre-processing code if possible?
thank you very much

Is there an easy way to generate new images using trained models with specific inputs?

I tried to run ./demo_lfw_trainCVAE.sh on my machine but it is giving an error. Assuming that it may be due to not having an appropriate GPU I rerun it with the flag that states not to use GPU:

th scripts/train_lfw_condVAE.lua --maxEpoch 100 --gpu -1

However I get an identical error:

miguel@whitney:~/eccv16_attr2img/eccv16_attr2img-master$ th scripts/train_lfw_condVAE.lua --maxEpoch 100 --gpu -1
THCudaCheck FAIL file=/tmp/luarocks_cutorch-scm-1-541/cutorch/lib/THC/THCGeneral.c line=70 error=30 : unknown error
/home/miguel/torch/install/bin/luajit: /home/miguel/torch/install/share/lua/5.1/trepl/init.lua:389: cuda runtime error (30) : unknown error at /tmp/luarocks_cutorch-scm-1-541/cutorch/lib/THC/THCGeneral.c:70
stack traceback:
[C]: in function 'error'
/home/miguel/torch/install/share/lua/5.1/trepl/init.lua:389: in function 'require'
scripts/train_lfw_condVAE.lua:2: in main chunk
[C]: in function 'dofile'
...guel/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
[C]: at 0x00405d50

I believe that may be related to the "require 'cutorch'" in train_lfw_condVAE.lua - why require cutorch if I stated not to use GPU?

Note that in order to avoid similar errors I was getting before I installed cutorch and cuda (following the instructions I found in various places), but I am still getting the error above.

Thank you.

Hi, I wonder how to get the corresponding attribute vector of a given image (one not in the valData) for testing, so I can just change one attribute while keep others intact.