dcgan_mnist

Deep Convolutional GAN (DC-GAN)

Source

Model Info

A DC-GAN is a direct extension of the GAN, except that it explicitly uses convolutional and transposed convolutional layers in the discriminator and generator, respectively. The discriminator is made up of strided convolutional layers, batch norm layers, and LeakyReLU activations. The generator is comprised of transposed convolutional layers, batch norm layers, and ReLU activations.

Training

cd vision/dcgan_mnist
julia --project dcgan_mnist.jl

Results

2000 training steps

5000 training steps

8000 training steps

9380 training steps

References

• Radford, A. et al.: Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks, http://arxiv.org/abs/1511.06434, (2015).

• pytorch.org/tutorials/beginner/dcgan_faces_tutorial