← Improved Training of Wasserstein GANs