TORE: Token Recycling in Vision Transformers for Efficient Active Visual Exploration

Jan Olszewski, Dawid Rymarczyk, Piotr Wójcik, Mateusz Pach, Bartosz Zieliński
11/26/2023
cs.LGcs.AIcs.CV

Official code: jano1906/TokenRecycling (0 stars, MIT, last commit 2026-03-13).

Abstract

Active Visual Exploration (AVE) optimizes the utilization of robotic resources in real-world scenarios by sequentially selecting the most informative observations. However, modern methods require a high computational budget due to processing the same observations multiple times through the autoencoder transformers. As a remedy, we introduce a novel approach to AVE called TOken REcycling (TORE). It divides the encoder into extractor and aggregator components. The extractor processes each observation separately, enabling the reuse of tokens passed to the aggregator. Moreover, to further reduce the computations, we decrease the decoder to only one block. Through extensive experiments, we demonstrate that TORE outperforms state-of-the-art methods while reducing computational overhead by up to 90\%.

View on arXivView PDF

Code Implementations(1)

jano1906/TokenRecyclingOfficial100%
MIT

Token Recycling in Vision Transformers for Efficient Active Visual Exploration

00Shell, PythonNov 22, 20247 months agoMIT

Similar Papers

Making Vision Transformers Efficient from A Token Sparsification View

82% match2 repos342023

Principles of Visual Tokens for Efficient Video Understanding

81% match1 repo2024

HiRED: Attention-Guided Token Dropping for Efficient Inference of High-Resolution Vision-Language Models

81% match1 repo562024

Efficient Visual Transformer by Learnable Token Merging

81% match1 repo52024

Efficient and Accurate Scene Text Recognition with Cascaded-Transformers

81% match2025

Streamline Without Sacrifice - Squeeze out Computation Redundancy in LMM

80% match1 repo202026

Cite this paper

@article{olszewski2023tore,
  title  = {TORE: Token Recycling in Vision Transformers for Efficient Active Visual Exploration},
  author = {Jan Olszewski and Dawid Rymarczyk and Piotr Wójcik and Mateusz Pach and Bartosz Zieliński},
  year   = {2023},
  eprint = {2311.15335},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2311.15335v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2311.15335.svg)](https://codeofpaper.com/paper/2311.15335)

Discussion