Enregistré dans:
Détails bibliographiques
Auteurs principaux: Nan, Xinyu, Mao, Lingtao, Dai, Huangyu, Zheng, Zexin, Sun, Xinyu, Liang, Zihan, Chen, Ben, Ding, Yuqing, Lei, Chenyi, Ou, Wenwu, Li, Han
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2511.15984
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866911276880363520
author Nan, Xinyu
Mao, Lingtao
Dai, Huangyu
Zheng, Zexin
Sun, Xinyu
Liang, Zihan
Chen, Ben
Ding, Yuqing
Lei, Chenyi
Ou, Wenwu
Li, Han
author_facet Nan, Xinyu
Mao, Lingtao
Dai, Huangyu
Zheng, Zexin
Sun, Xinyu
Liang, Zihan
Chen, Ben
Ding, Yuqing
Lei, Chenyi
Ou, Wenwu
Li, Han
contents Achieving visual semantic understanding requires a unified framework that simultaneously handles object detection, category prediction, and attribute recognition. However, current advanced approaches rely on global similarity and struggle to capture fine-grained category distinctions and category-specific attribute diversity, especially in large-scale e-commerce scenarios. To overcome these challenges, we introduce a detection-guided generative framework that predicts hierarchical category and attribute tokens. For each detected object, we extract refined ROI-level features and employ a BART-based generator to produce semantic tokens in a coarse-to-fine sequence covering category hierarchies and property-value pairs, with support for property-conditioned attribute recognition. Experiments on both large-scale proprietary e-commerce datasets and open-source datasets demonstrate that our approach significantly outperforms existing similarity-based pipelines and multi-stage classification systems, achieving stronger fine-grained recognition and more coherent unified inference.
format Preprint
id arxiv_https___arxiv_org_abs_2511_15984
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle UniDGF: A Unified Detection-to-Generation Framework for Hierarchical Object Visual Recognition
Nan, Xinyu
Mao, Lingtao
Dai, Huangyu
Zheng, Zexin
Sun, Xinyu
Liang, Zihan
Chen, Ben
Ding, Yuqing
Lei, Chenyi
Ou, Wenwu
Li, Han
Computer Vision and Pattern Recognition
Achieving visual semantic understanding requires a unified framework that simultaneously handles object detection, category prediction, and attribute recognition. However, current advanced approaches rely on global similarity and struggle to capture fine-grained category distinctions and category-specific attribute diversity, especially in large-scale e-commerce scenarios. To overcome these challenges, we introduce a detection-guided generative framework that predicts hierarchical category and attribute tokens. For each detected object, we extract refined ROI-level features and employ a BART-based generator to produce semantic tokens in a coarse-to-fine sequence covering category hierarchies and property-value pairs, with support for property-conditioned attribute recognition. Experiments on both large-scale proprietary e-commerce datasets and open-source datasets demonstrate that our approach significantly outperforms existing similarity-based pipelines and multi-stage classification systems, achieving stronger fine-grained recognition and more coherent unified inference.
title UniDGF: A Unified Detection-to-Generation Framework for Hierarchical Object Visual Recognition
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.15984