kregmi / cross-view-image-synthesis Goto Github PK
View Code? Open in Web Editor NEW[CVPR 2018] Cross-View Image Synthesis using Conditional GANs, [CVIU 2019] Cross-view image synthesis using geometry-guided conditional GANs
[CVPR 2018] Cross-View Image Synthesis using Conditional GANs, [CVIU 2019] Cross-view image synthesis using geometry-guided conditional GANs
Dear author,
Thanks for your contribution!
I'm wondering why there should be 128 instead of 255.
The value of the image should be in range [0, 1] or [0, 255], instead of [-1, 1]
When I am using the following command:
DATA_ROOT=/data/dayton name=dayton_seq_64 which_direction=g2a phase=train batchSize=16 loadSize=72 fineSize=64 niter=100 th train_seq.lua
I hit the following error:
/data/torch/install/bin/lua: /data/torch/install/share/lua/5.2/cudnn/init.lua:166: Error in CuDNN: CUDNN_STATUS_BAD_PARAM (cudnnGetConvolutionNdForwardOutputDim) stack traceback: [C]: in function 'error' /data/torch/install/share/lua/5.2/cudnn/init.lua:166: in function 'errcheck' ...torch/install/share/lua/5.2/cudnn/SpatialConvolution.lua:139: in function 'createIODescriptors' ...torch/install/share/lua/5.2/cudnn/SpatialConvolution.lua:177: in function <...torch/install/share/lua/5.2/cudnn/SpatialConvolution.lua:175> (...tail calls...) /data/torch/install/share/lua/5.2/nngraph/gmodule.lua:345: in function 'neteval' /data/torch/install/share/lua/5.2/nngraph/gmodule.lua:380: in function </data/torch/install/share/lua/5.2/nngraph/gmodule.lua:300> (...tail calls...) train_seq.lua:293: in function 'createRealFake' train_seq.lua:490: in main chunk [C]: in function 'dofile' .../torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk [C]: in ?
Do you have any suggestions? Thanks.
Can you share pretrained model on CVUSA dataset?
Thanks for your paper and code!
I have a problem here, which confused me a lot.
KL (P || Q) = E_P[log(P/Q)] = \sum_i [p_i * (log p_i - log q_i]
So why add an np.exp
hehre
Thank you for sharing this amazing code.
As a strong baseline, could you mind also sharing the evaluation codes of KL and Sharpness Difference?
/home/shon/torch/install/bin/luajit: test_fork.lua:168: bad argument #1 to 'output' (/home/shon/cross-view-image-synthesis/results/sample_images/35_net_G_sample/index.html: No such file or directory)
stack traceback:
[C]: in function 'output'
test_fork.lua:168: in main chunk
[C]: in function 'dofile'
...shon/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
[C]: at 0x00405d50
I have this problem, how can I solve it? Thank you.
Can you kindly share the code of your extension paper "Cross-view image synthesis using geometry-guided conditional GANs"?
In your paper, you use 100 sets of training set images, but I found there are 200 sets of training set in SVA dataset. Can you share with us which part did you use?
Can you provide X-Seq training code?
Can you share the code for evaluation scores of segmentation maps?
@kregmi What kind of segmentation map are you used in CVUSA dataset? The color one or the black one?
If you used the color one, can you share the code for converting the back segmentation map to the color segmentation map?
Thank you in advance.
Could you mind sharing pre-trained models for 64*64 experiments?
/home/yep/torch/install/bin/luajit: /home/yep/torch/install/share/lua/5.1/torch/File.lua:343: unknown Torch class <torch.CudaTensor>
For compute_accuracies.py, I just want to make sure, is the
final result = total match found for synthesized / total images into consideration?
How to generate the segmentation map on X-Seq model?
Hi @kregmi
It's really interesting work. Since my environment is not suitable for installing Torch
, I am writing to ask kindly, is there any possibility for you to share the generated a2g
images on CVUSA and Dayton dataset? It would be so helpful for me and I really appreciate it.
My problems with your provided models in a docker image are :
(1) The CUDA driver version is insufficient for CUDA running time
(2) When I turn to test the models with CPU only, I failed to test it on CPU with the trained model from GPU.
Also, it would be very nice if you also can provide me with some suggestions on how to solve these problems
Many thanks,
Can you provide the splits for Dayton dataset?
Could you share the Crossview USA (CVUSA) to me?
Hi,
I've been trying to test/reproduce the model on Google Colab, but not sure what am I doing wrong, the link of the notebook with the steps I've taken can be found here.
The summary of the steps is:
checkpoints
directory inside the repository.DATA_ROOT=./datasets/AB_AsBs name=sample_images which_direction=a2g phase=sample which_epoch=35 th test_fork.lua
The output error can be found in the last cell in the notebook but here is a screenshot
Any help would be appreciated, thanks.
Can you share the training/testing split and segmentation maps of the SVA dataset?
And can you share the image ID of Fig 7 in your extension paper for comparison?
Thank you in advance!
DATA_ROOT=./datasets/AB_AsBs name=a2g_fork which_direction=a2g phase=test which_epoch=35 th test_fork.lua
{
input_nc : 3
results_dir : "./results/"
name : "a2g_fork"
batchSize : 10
phase : "test"
fineSize : 256
aspect_ratio : 1
how_many : "all"
gpu : 1
nThreads : 1
DATA_ROOT : "./datasets/AB_AsBs"
serial_batch_iter : 1
output_nc_seg : 3
which_epoch : 35
loadSize : 256
cudnn : 1
serial_batches : 1
which_direction : "a2g"
display : 0
output_nc : 3
preprocess : "regular"
checkpoints_dir : "./checkpoints"
display_id : 200
flip : 0
}
Random Seed: 4974
#threads...1
Starting donkey with id: 1 seed: 4975
table: 0x41bfb1c8
./datasets/AB_AsBs
trainCache /home/csdept/projects/cross-view-image-synthesis/cache/_home_csdept_projects_cross-view-image-synthesis_datasets_AB_AsBs_test_trainCache.t7
Creating train metadata
serial batch:, 1
table: 0x40732bf0
running "find" on each class directory, and concatenate all those filenames into a single file containing all image paths for a given class
now combine all the files to a single large file
load the large concatenated list of sample paths to self.imagePath
cmd..wc -L '/tmp/lua_jMFhFH' |cut -f1 -d' '
8 samples found........................... 0/8 .........................................] ETA: 0ms | Step: 0ms
Updating classList and imageClass appropriately
[======================================== 1/1 ========================================>] Tot: 0ms | Step: 0ms
Cleaning up temporary files
Dataset Size: 8
checkpoints_dir ./checkpoints
nn.gModule
No outputs generated.
Could you share the segmentation maps of the Dayton dataset generated by RefineNet?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.