Knowledge Commons of Institute of Automation,CAS
Robust Feature Rectification of Pretrained Vision Models for Object Recognition | |
Zhou SC(周圣超)1,2; Meng GF(孟高峰)1,2,3; Zhang ZX(张兆翔)1,2,3; Xu YD(徐亦达)4; Xiang SM(向世明)1,2 | |
2023-06-26 | |
会议名称 | Proceedings of the AAAI Conference on Artificial Intelligence |
会议日期 | 2023-2-7 |
会议地点 | 美国华盛顿 |
摘要 | Pretrained vision models for object recognition often suffer a dramatic performance drop with degradations unseen during training. In this work, we propose a RObust FEature Rectification module (ROFER) to improve the performance of pretrained models against degradations. Specifically, ROFER first estimates the type and intensity of the degradation that corrupts the image features. Then, it leverages a Fully Convolutional Network (FCN) to rectify the features from the degradation by pulling them back to clear features. ROFER is a general-purpose module that can address various degradations simultaneously, including blur, noise, and low contrast. Besides, it can be plugged into pretrained models seamlessly to rectify the degraded features without retraining the whole model. Furthermore, ROFER can be easily extended to address composite degradations by adopting a beam search algorithm to find the composition order. Evaluations on CIFAR-10 and Tiny-ImageNet demonstrate that the accuracy of ROFER is 5% higher than that of SOTA methods on different degradations. With respect to composite degradations, ROFER improves the accuracy of a pretrained CNN by 10% and 6% on CIFAR-10 and Tiny-ImageNet respectively. |
语种 | 英语 |
七大方向——子方向分类 | 目标检测、跟踪与识别 |
国重实验室规划方向分类 | 视觉信息处理 |
是否有论文关联数据集需要存交 | 否 |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/56491 |
专题 | 多模态人工智能系统全国重点实验室 |
通讯作者 | Meng GF(孟高峰) |
作者单位 | 1.中国科学院自动化研究所 2.中国科学院大学人工智能学院 3.中国科学院香港创新研究院人工智能与机器人创新研究中心 4.香港浸会大学 |
第一作者单位 | 中国科学院自动化研究所 |
通讯作者单位 | 中国科学院自动化研究所 |
推荐引用方式 GB/T 7714 | Zhou SC,Meng GF,Zhang ZX,et al. Robust Feature Rectification of Pretrained Vision Models for Object Recognition[C],2023. |
条目包含的文件 | 下载所有文件 | |||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
aaai23.pdf(1238KB) | 会议论文 | 开放获取 | CC BY-NC-SA | 浏览 下载 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论