Jingcheng Ke, Dele Wang, Jun-Cheng Chen, I-Hong Jhuo, Chia-Wen Lin, Yen-Yu Lin
{"title":"通过以表达为导向的动态门控和回归,让基于图表的引用表达理解再创辉煌","authors":"Jingcheng Ke, Dele Wang, Jun-Cheng Chen, I-Hong Jhuo, Chia-Wen Lin, Yen-Yu Lin","doi":"arxiv-2409.03385","DOIUrl":null,"url":null,"abstract":"One common belief is that with complex models and pre-training on large-scale\ndatasets, transformer-based methods for referring expression comprehension\n(REC) perform much better than existing graph-based methods. We observe that\nsince most graph-based methods adopt an off-the-shelf detector to locate\ncandidate objects (i.e., regions detected by the object detector), they face\ntwo challenges that result in subpar performance: (1) the presence of\nsignificant noise caused by numerous irrelevant objects during reasoning, and\n(2) inaccurate localization outcomes attributed to the provided detector. To\naddress these issues, we introduce a plug-and-adapt module guided by\nsub-expressions, called dynamic gate constraint (DGC), which can adaptively\ndisable irrelevant proposals and their connections in graphs during reasoning.\nWe further introduce an expression-guided regression strategy (EGR) to refine\nlocation prediction. Extensive experimental results on the RefCOCO, RefCOCO+,\nRefCOCOg, Flickr30K, RefClef, and Ref-reasoning datasets demonstrate the\neffectiveness of the DGC module and the EGR strategy in consistently boosting\nthe performances of various graph-based REC methods. Without any pretaining,\nthe proposed graph-based method achieves better performance than the\nstate-of-the-art (SOTA) transformer-based methods.","PeriodicalId":501480,"journal":{"name":"arXiv - CS - Multimedia","volume":"7 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Make Graph-based Referring Expression Comprehension Great Again through Expression-guided Dynamic Gating and Regression\",\"authors\":\"Jingcheng Ke, Dele Wang, Jun-Cheng Chen, I-Hong Jhuo, Chia-Wen Lin, Yen-Yu Lin\",\"doi\":\"arxiv-2409.03385\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"One common belief is that with complex models and pre-training on large-scale\\ndatasets, transformer-based methods for referring expression comprehension\\n(REC) perform much better than existing graph-based methods. We observe that\\nsince most graph-based methods adopt an off-the-shelf detector to locate\\ncandidate objects (i.e., regions detected by the object detector), they face\\ntwo challenges that result in subpar performance: (1) the presence of\\nsignificant noise caused by numerous irrelevant objects during reasoning, and\\n(2) inaccurate localization outcomes attributed to the provided detector. To\\naddress these issues, we introduce a plug-and-adapt module guided by\\nsub-expressions, called dynamic gate constraint (DGC), which can adaptively\\ndisable irrelevant proposals and their connections in graphs during reasoning.\\nWe further introduce an expression-guided regression strategy (EGR) to refine\\nlocation prediction. Extensive experimental results on the RefCOCO, RefCOCO+,\\nRefCOCOg, Flickr30K, RefClef, and Ref-reasoning datasets demonstrate the\\neffectiveness of the DGC module and the EGR strategy in consistently boosting\\nthe performances of various graph-based REC methods. Without any pretaining,\\nthe proposed graph-based method achieves better performance than the\\nstate-of-the-art (SOTA) transformer-based methods.\",\"PeriodicalId\":501480,\"journal\":{\"name\":\"arXiv - CS - Multimedia\",\"volume\":\"7 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-05\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Multimedia\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.03385\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Multimedia","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.03385","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Make Graph-based Referring Expression Comprehension Great Again through Expression-guided Dynamic Gating and Regression
One common belief is that with complex models and pre-training on large-scale
datasets, transformer-based methods for referring expression comprehension
(REC) perform much better than existing graph-based methods. We observe that
since most graph-based methods adopt an off-the-shelf detector to locate
candidate objects (i.e., regions detected by the object detector), they face
two challenges that result in subpar performance: (1) the presence of
significant noise caused by numerous irrelevant objects during reasoning, and
(2) inaccurate localization outcomes attributed to the provided detector. To
address these issues, we introduce a plug-and-adapt module guided by
sub-expressions, called dynamic gate constraint (DGC), which can adaptively
disable irrelevant proposals and their connections in graphs during reasoning.
We further introduce an expression-guided regression strategy (EGR) to refine
location prediction. Extensive experimental results on the RefCOCO, RefCOCO+,
RefCOCOg, Flickr30K, RefClef, and Ref-reasoning datasets demonstrate the
effectiveness of the DGC module and the EGR strategy in consistently boosting
the performances of various graph-based REC methods. Without any pretaining,
the proposed graph-based method achieves better performance than the
state-of-the-art (SOTA) transformer-based methods.