MixNMatch: Multifactor Disentanglement and Encoding for Conditional Image Generation
This addresses the challenge of controllable image synthesis for computer vision applications, though it builds incrementally on existing methods like FineGAN.
The paper tackles the problem of disentangling and encoding multiple factors like background, pose, shape, and texture from real images with minimal supervision for conditional image generation, achieving accurate mix-and-match generation in applications such as sketch2color and img2gif.
We present MixNMatch, a conditional generative model that learns to disentangle and encode background, object pose, shape, and texture from real images with minimal supervision, for mix-and-match image generation. We build upon FineGAN, an unconditional generative model, to learn the desired disentanglement and image generator, and leverage adversarial joint image-code distribution matching to learn the latent factor encoders. MixNMatch requires bounding boxes during training to model background, but requires no other supervision. Through extensive experiments, we demonstrate MixNMatch's ability to accurately disentangle, encode, and combine multiple factors for mix-and-match image generation, including sketch2color, cartoon2img, and img2gif applications. Our code/models/demo can be found at https://github.com/Yuheng-Li/MixNMatch