.png){kind=icon}
返回
Complex image classification by feature inference
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2023.126231.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Image classification is a fundamental task in image processing. Despite the long time research, there are still many challenging problems to be solved. In this study, we introduce the problem of complex image classification. Images in realistic scenarios are complex and classifying samples directly are not always the right way even the performance is high. To address the issue, we propose a novel classification schema where classification is combined with image inpainting. There are two models including one inference network and one classification network in the proposed schema. The masked content that is specified as occlusions and interferences is inpainted by feature inference. The inference network inpaint the image with a mask and the network classifies the inferred image. We apply the proposed schema to existing classification models including AlexNet, GoogleNet, Inceptionv3, ResNet50, and EfficientNetb7 and specific datasets including ImageNet, PlantCLEF, and CUB-200. Despite the simplicity, experimental results show that it significantly improves the performance of complex classifications.(c) 2023 Elsevier B.V. All rights reserved.
Keyword:
Image classification
Complex sample
Image inference
Visual attention
期刊
IF:
6.5
论文数:
2.5W
被引数:
6.5W
机构
引用论文
Hidden trauma victims: Understanding and preventing traumatic stress in mental health professionals
.png){kind=icon}
PREAI
分享收藏Assessment of Impact on Health of Children Working in the Garbage Dumping Site in Dhaka, Bangladesh
OAAI
OAAI分享收藏
分享收藏分享收藏
查看更多内容