SequenceGan: Text to Image Synthesis with Sequence Models and Gans

Generative Adversarial Nets are one of the most popular generative frameworks. In our work, we introduce the SequenceGAN, a method that can generate images based on a given caption by supplying the conditional sequential text input to the generator and the discriminator. Unlike other conditional methods, SequenceGAN uses recurrent layers for better context understanding. We also demonstrated SequenceGANs performance by applying it to the MNIST and Flickr 8k datasets.