计算机科学 ›› 2023, Vol. 50 ›› Issue (11A): 220800218-6.doi: 10.11896/jsjkx.220800218
袁振, 刘进锋
YUAN Zhen, LIU Jinfeng
摘要: 垃圾分类是保护环境,提高资源利用率的有效措施。近年来,深度学习以其引入强大的建模能力在各个领域都有成功的表现,利用深度学习进行垃圾分类也成为一个新兴方向。垃圾分类数据集大多存在类别图片数量不均衡(长尾分布)的问题。针对垃圾分类中的长尾数据集TrashNet,提出了一种基于因果推理分类方法。该方法通过引入因果推理找出由输入样本引起的直接因果效应,缓解了TrashNet数据集的长尾问题。在训练中采用了迁移学习的方法,减少了训练参数量,并在其中使用了因果干预与反事实推理,进行了去混淆训练。对所提出的方法进行了类激活图(Class Activation Map)验证,实验结果表明,所提出的模型有较好的特征提取能力。模型对TrashNet数据集中的尾部类有较优的识别效果,并在其上取得了94.23%的最好精度。
中图分类号:
[1]PRASANNA A,KAUSHAL S V,MAHALAKSHMI P,et al.Survey on identification and classification of waste for efficient disposal and recycling[J].International Journal of Engineering &Technology,2018,7(2.8):520-523. [2]CUDJOE D,ZHU B,NKETIAH E,et al.The potential energy and environmental benefits of global recyclable resources[J].Science of The Total Environment,2021,798:149258. [3]LECUN Y,BENGIO Y,HINTON G J N.Deep learning[J].Nature,2015,521(7553):436-44. [4]YANG THUNG G.Classification of trash for recyclability status[J].CS229 Project Report,2016,2016(1):3. [5]ARAL R A,KESKIN Ş R,KAYA M,et al.Classification oftrashnet dataset based on deep learning models[C]//2018 IEEE International Conference on Big Data(Big Data).New York:IEEE,2018:2058-2062. [6]SHI C,XIAR,WANG L.A novel multi-branch channel expansion network for garbage image classification[J].Ieee Access,2020,8:154436-154452. [7]VO A H,VO M T,LE T.A novel framework for trash classifi-cation using deep transfer learning[J].IEEE Access,2019,7:178631-178639. [8]ZHOU B,KHOSLA A,LAPEDRIZA A,et al.Learning deepfeatures for discriminative localization[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.New York:IEEE Press,2016:2921-2929. [9]BIRCANOĞLU C,ATAY M,BEŞER F,et al.RecycleNet:Intelligent waste sorting using deep neural networks[C]//2018 Innovations in Intelligent Systems and Applications(INISTA).New York:IEEE,2018:1-7. [10]RUIZ V,SÁNCHEZ Á,VÉLEZ J F,et al.Automatic image-based waste classification[C]//International Work-Conference on the Interplay Between Natural and Artificial Computation.Cham:Springer,2019:422-431. [11]PEARL J.Causal diagrams for empirical research[J].Biometrika,1995,82(4):669-688. [12]PEARL J.Direct and indirect effects[M]//Probabilistic andCausal Inference:The Works of Judea Pearl.2022:373-392. [13]TANG K,HUANG J,ZHANG H.Long-tailed classification by keeping the good and removing the bad momentum causal effect[J].Advances in Neural Information Processing Systems,2020,33:1513-1524. [14]VANDERWEELE T J.A three-way decomposition of a totaleffect into direct,indirect,and interactive effects[J].Epidemio-logy(Cambridge,Mass.),2013,24(2):224. [15]GLYMOUR M,PEARL J,JEWELL N P.Causal inference in statistics:A primer[M].John Wiley & Sons,2016. [16]LECUN Y,CHOPRA S,HADSELL R,et al.A tutorial on energy-based learning[C]//Predicting Structured Data.2006. [17]VASWANI A,SHAZEER N,PARMAR N,et al.Attention isall you need[J].Advances in Meural Information Processing Systems,2017,30. [18]PEARL J.On the Consistency Rule in Causal Inference:“Axiom,Definition,Assumption,or Theorem?”[J].Epidemiology,2011,22(2):285-285. [19]XIE S,GIRSHICK R,DOLLÁR P,et al.Aggregated residualtransformations for deep neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.New York:IEEE Press,2017:1492-1500. |
|