文献类型：期刊文献

中文题名：Causal reasoning in typical computer vision tasks

英文题名：Causal reasoning in typical computer vision tasks

作者：Zhang, Kexuan[1];Sun, Qiyu[1];Zhao, Chaoqiang[2,3];Tang, Yang[1]

机构：[1]East China Univ Sci & Technol, Key Lab Adv Control & Optimizat Chem Proc, Minist Educ, Shanghai 200237, Peoples R China;[2]Natl Key Lab Air Based Informat Percept & Fus, Luoyang 471000, Peoples R China;[3]Luoyang Inst Electro Opt Equipment Avic, Luoyang 471000, Peoples R China

年份：2024

卷号：67

期号：1

起止页码：105

中文期刊名：Science China(Technological Sciences)

外文期刊名：SCIENCE CHINA-TECHNOLOGICAL SCIENCES

收录：CSTPCD;;EI(收录号：20240315395932);Scopus;WOS:【SCI-EXPANDED(收录号:WOS:001126614800002)】；CSCD:【CSCD2023_2024】；PubMed;

基金：This work was supported by the National Natural Science Foundation of China (Grant Nos. 62233005 and 62293502), the Programme of Introducing Talents of Discipline to Universities (the 111 Project, Grant No. B17017),the Fundamental Research Funds for the Central Universities (Grant No.222202317006), and Shanghai AI Lab.

语种：英文

中文关键词：causal reasoning;computer vision tasks;vision-language tasks;semantic segmentation;object detection

外文关键词：causal reasoning; computer vision tasks; vision-language tasks; semantic segmentation; object detection

摘要：Deep learning has revolutionized the field of artificial intelligence.Based on the statistical correlations uncovered by deep learning-based methods,computer vision tasks,such as autonomous driving and robotics,are growing rapidly.Despite being the basis of deep learning,such correlation strongly depends on the distribution of the original data and is susceptible to uncontrolled factors.Without the guidance of prior knowledge,statistical correlations alone cannot correctly reflect the essential causal relations and may even introduce spurious correlations.As a result,researchers are now trying to enhance deep learningbased methods with causal theory.Causal theory can model the intrinsic causal structure unaffected by data bias and effectively avoids spurious correlations.This paper aims to comprehensively review the existing causal methods in typical vision and visionlanguage tasks such as semantic segmentation,object detection,and image captioning.The advantages of causality and the approaches for building causal paradigms will be summarized.Future roadmaps are also proposed,including facilitating the development of causal theory and its application in other complex scenarios and systems.
Deep learning has revolutionized the field of artificial intelligence. Based on the statistical correlations uncovered by deep learning-based methods, computer vision tasks, such as autonomous driving and robotics, are growing rapidly. Despite being the basis of deep learning, such correlation strongly depends on the distribution of the original data and is susceptible to uncontrolled factors. Without the guidance of prior knowledge, statistical correlations alone cannot correctly reflect the essential causal relations and may even introduce spurious correlations. As a result, researchers are now trying to enhance deep learning-based methods with causal theory. Causal theory can model the intrinsic causal structure unaffected by data bias and effectively avoids spurious correlations. This paper aims to comprehensively review the existing causal methods in typical vision and vision-language tasks such as semantic segmentation, object detection, and image captioning. The advantages of causality and the approaches for building causal paradigms will be summarized. Future roadmaps are also proposed, including facilitating the development of causal theory and its application in other complex scenarios and systems.