Saved in:
Bibliographic Details
Main Authors: Chen, Delong, Cahyawijaya, Samuel, Liu, Jianfeng, Wang, Baoyuan, Fung, Pascale
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2402.14327
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929756484665344
author Chen, Delong
Cahyawijaya, Samuel
Liu, Jianfeng
Wang, Baoyuan
Fung, Pascale
author_facet Chen, Delong
Cahyawijaya, Samuel
Liu, Jianfeng
Wang, Baoyuan
Fung, Pascale
contents Patch-based image tokenization ignores the morphology of the visual world, limiting effective and efficient learning of image understanding. Inspired by subword tokenization, we introduce subobject-level adaptive token segmentation and explore several approaches, including superpixel, SAM, and a proposed Efficient and PanOptiC (EPOC) image tokenizer. Our EPOC combines boundary detection -- a simple task that can be handled well by a compact model -- with watershed segmentation, which inherently guarantees no pixels are left unsegmented. Intrinsic evaluations across 5 datasets demonstrate that EPOC's segmentation aligns well with human annotations of both object- and part-level visual morphology, producing more monosemantic tokens and offering substantial efficiency advantages. For extrinsic evaluation, we designed a token embedding that handles arbitrary-shaped tokens, and trained VLMs with different tokenizers on 4 datasets of object recognition and detailed captioning. The results reveal that subobject tokenization enables faster convergence and better generalization while using fewer visual tokens.
format Preprint
id arxiv_https___arxiv_org_abs_2402_14327
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Subobject-level Image Tokenization
Chen, Delong
Cahyawijaya, Samuel
Liu, Jianfeng
Wang, Baoyuan
Fung, Pascale
Computer Vision and Pattern Recognition
Computation and Language
Patch-based image tokenization ignores the morphology of the visual world, limiting effective and efficient learning of image understanding. Inspired by subword tokenization, we introduce subobject-level adaptive token segmentation and explore several approaches, including superpixel, SAM, and a proposed Efficient and PanOptiC (EPOC) image tokenizer. Our EPOC combines boundary detection -- a simple task that can be handled well by a compact model -- with watershed segmentation, which inherently guarantees no pixels are left unsegmented. Intrinsic evaluations across 5 datasets demonstrate that EPOC's segmentation aligns well with human annotations of both object- and part-level visual morphology, producing more monosemantic tokens and offering substantial efficiency advantages. For extrinsic evaluation, we designed a token embedding that handles arbitrary-shaped tokens, and trained VLMs with different tokenizers on 4 datasets of object recognition and detailed captioning. The results reveal that subobject tokenization enables faster convergence and better generalization while using fewer visual tokens.
title Subobject-level Image Tokenization
topic Computer Vision and Pattern Recognition
Computation and Language
url https://arxiv.org/abs/2402.14327