Most existing image tokenizers encode images into a fixed number of tokens or patches, overlooking the inherent variability in image complexity. To address this, we introduce Content-Adaptive Tokenizer (CAT), which dynamically adjusts representation capacity based on the image content and encodes simpler images into fewer tokens. We design a caption-based evaluation system that leverages large language models (LLMs) to predict content complexity and determine the optimal compression ratio for a given image, taking into account factors critical to human perception. Trained on images with diverse compression ratios, CAT demonstrates robust performance in image reconstruction. We also utilize its variable-length latent representations to train Diffusion Transformers (DiTs) for ImageNet generation. By optimizing token allocation, CAT improves the FID score over fixed-ratio baselines trained with the same flops and boosts the inference throughput by 18.5%.
@article{arxiv.2501.03120,
title = {CAT: Content-Adaptive Image Tokenization},
author = {Junhong Shen and Kushal Tirumala and Michihiro Yasunaga and Ishan Misra and Luke Zettlemoyer and Lili Yu and Chunting Zhou},
journal= {arXiv preprint arXiv:2501.03120},
year = {2025}
}