IMAGINE: Image Synthesis by Image-Guided Model Inversion

TL;DR

IMAGINE is a novel image synthesis method that generates diverse, high-quality images from a single sample by leveraging pre-trained classifiers and adversarial training, offering semantic control without generator training.

Contribution

The paper introduces IMAGINE, a new inversion-based approach that synthesizes images from one sample using classifier features and adversarial training, without needing to train a generator.

Findings

Performs favorably against state-of-the-art methods in multiple domains.

Produces realistic and diverse images from a single sample.

Enables semantic control over generated images.

Abstract

We introduce an inversion based method, denoted as IMAge-Guided model INvErsion (IMAGINE), to generate high-quality and diverse images from only a single training sample. We leverage the knowledge of image semantics from a pre-trained classifier to achieve plausible generations via matching multi-level feature representations in the classifier, associated with adversarial training with an external discriminator. IMAGINE enables the synthesis procedure to simultaneously 1) enforce semantic specificity constraints during the synthesis, 2) produce realistic images without generator training, and 3) give users intuitive control over the generation process. With extensive experimental results, we demonstrate qualitatively and quantitatively that IMAGINE performs favorably against state-of-the-art GAN-based and inversion-based methods, across three different image domains (i.e., objects, scenes, and textures).