text-to-image

Why we included this project

This is a compact, readable implementation of one of the earliest text-to-image approaches. It follows the GAN-CLS method from the Generative Adversarial Text-to-Image Synthesis paper, pairing Skip-Thought caption embeddings with a DCGAN backbone so a short flower description becomes a 64x64 image. The repo includes training and inference scripts, a data downloader, and a pretrained checkpoint for the flowers dataset, so you can go from captions to generated samples without rerunning a full training pass. It is clearly a research artifact rather than a maintained product: the pipeline depends on Python 2.7 and a TensorFlow-era toolchain. If you are working through the original algorithm, or want a simple baseline to compare against modern text-to-image models, studying this code is likely to be more useful than trying to deploy it.