StyleFlow: Attribute-conditioned Exploration of StyleGAN-Generated Images using Conditional Continuous Normalizing Flows

High-quality, diverse, and photorealistic images can now be generated by unconditional GANs (e.g., StyleGAN). However, limited options exist to control the generation process using (semantic) attributes, while still preserving quality of output. Further, due entangled nature GAN latent space, performing edits along one attribute easily result in unwanted changes other attributes. In this paper, context conditional exploration spaces, we investigate two sub-problems attribute-conditioned sampling attribute-controlled editing. We present StyleFlow as a simple, effective, robust solution both formulating an instance continuous normalizing flows space conditioned features. evaluate our method face car StyleGAN, demonstrate fine-grained disentangled various attributes on real photographs StyleGAN images. For example, for faces, vary camera pose, illumination variation, expression, facial hair, gender, age. Finally, via extensive qualitative quantitative comparisons, superiority concurrent works.