Title: CAT: Content-Adaptive Image Tokenization
Authors: Junhong Shen, Kushal Tirumala, Michihiro Yasunaga, Ishan Misra, Luke Zettlemoyer, Lili Yu, Chunting Zhou
Published: 6th January 2025 (Monday) @ 16:28:47
Link: http://arxiv.org/abs/2501.03120v1
Abstract
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%.
Found thanks to this tweet from Simone Scardapane, retweeted by Ishan Misra.
Made me think of Principles of Visual Tokens for Efficient Video Understanding