Simulated+Unsupervised (S+U) learning in TensorFlow

NYU Hand Dataset

Another TensorFlow implementation of Learning from Simulated and Unsupervised Images through Adversarial Training.

Thanks to TaeHoon Kim, I was able to run simGAN that generates refined synthetic eye dataset.
This is just another version of his code that can generate NYU hand datasets.

The structure of the refiner/discriminator networks are changed as it is described in the Apple paper.
The only code added in this version is ./data/hand_data.py.
Rest of the code runs in the same way as the original version.
To set up the environment(or to run UnityEyes dataset), please follow instructions in this link.

###Notes -NYU hand dataset is preprocessed(e.g. background removed)
-Image size set to 128x128
-Buffer/Batch size was reduced due to memory issues
-Local adversarial loss implemented with window size 2x2 but produces artifacts
*(will test further with smaller window size)
-May have to change the size of the discriminator network since the size of the patches are smaller than the original input images

##Results

Given these synthetic images,

###Test 1

'lambda=0.5' with 'optimizer=sgd'
After 4000 steps.

After ~20k steps

###Test 2

'lambda=0.1' with 'optimizer=sgd' after 20k steps.

###Test 3

'lambda=1.0' with 'optimizer=sgd' after 20k steps.

Author

Seung Shin / @shinseung428