# ImagePiece: Content-aware Re-tokenization for Efficient Image Recognition

Source: [ImagePiece: Content-aware Re-tokenization for Efficient Image Recognition](<https://www.sota2.com/research/paper/68fa765c1e209131cea6c9d6>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T03:52:28.667Z

Authors: Seungdong Yoa, Seungjun Lee, Hyeseung Cho, Bumsoo Kim, Woohyung Lim

Published: 2024-12-21T05:38:20.000Z

## Abstract

Vision Transformers (ViTs) have achieved remarkable success in various computer vision tasks. However, ViTs have a huge computational cost due to their inherent reliance on multi-head self-attention (MHSA), prompting efforts to accelerate ViTs for practical applications. To this end, recent works aim to reduce the number of tokens, mainly focusing on how to effectively prune or merge them. Nevertheless, since ViT tokens are generated from non-overlapping grid patches, they usually do not convey sufficient semantics, making it incompatible with efficient ViTs. To address this, we propose ImagePiece, a novel re-tokenization strategy for Vision Transformers. Following the MaxMatch strategy of NLP tokenization, ImagePiece groups semantically insufficient yet locally coherent tokens until they convey meaning. This simple retokenization is highly compatible with previous token reduction methods, being able to drastically narrow down relevant tokens, enhancing the inference speed of DeiT-S by 54% (nearly 1.5$\\times$ faster) while achieving a 0.39% improvement in ImageNet classification accuracy. For hyper-speed inference scenarios (with 251% acceleration), our approach surpasses other baselines by an accuracy over 8%.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2412.16491>)

[PDF](<https://arxiv.org/pdf/2412.16491.pdf>)
