English

CycleVAR: Repurposing Autoregressive Model for Unsupervised One-Step Image Translation

Computer Vision and Pattern Recognition 2025-07-08 v2

Abstract

The current conditional autoregressive image generation methods have shown promising results, yet their potential remains largely unexplored in the practical unsupervised image translation domain, which operates without explicit cross-domain correspondences. A critical limitation stems from the discrete quantization inherent in traditional Vector Quantization-based frameworks, which disrupts gradient flow between the Variational Autoencoder decoder and causal Transformer, impeding end-to-end optimization during adversarial training in image space. To tackle this issue, we propose using Softmax Relaxed Quantization, a novel approach that reformulates codebook selection as a continuous probability mixing process via Softmax, thereby preserving gradient propagation. Building upon this differentiable foundation, we introduce CycleVAR, which reformulates image-to-image translation as image-conditional visual autoregressive generation by injecting multi-scale source image tokens as contextual prompts, analogous to prefix-based conditioning in language models. CycleVAR exploits two modes to generate the target image tokens, including (1) serial multi-step generation, enabling iterative refinement across scales, and (2) parallel one-step generation synthesizing all resolution outputs in a single forward pass. Experimental findings indicate that the parallel one-step generation mode attains superior translation quality with quicker inference speed than the serial multi-step mode in unsupervised scenarios. Furthermore, both quantitative and qualitative results indicate that CycleVAR surpasses previous state-of-the-art unsupervised image translation models, \textit{e}.\textit{g}., CycleGAN-Turbo.

Keywords

Cite

@article{arxiv.2506.23347,
  title  = {CycleVAR: Repurposing Autoregressive Model for Unsupervised One-Step Image Translation},
  author = {Yi Liu and Shengqian Li and Zuzeng Lin and Feng Wang and Si Liu},
  journal= {arXiv preprint arXiv:2506.23347},
  year   = {2025}
}

Comments

Accepted to ICCV 2025. Code available at: https://github.com/IamCreateAI/CycleVAR

R2 v1 2026-07-01T03:38:40.261Z