Knowledge Commons of Institute of Automation,CAS
A Closer Look at Self-Supervised Lightweight Vision Transformers | |
Wang, Shaoru1,2; Gao, Jin1,2; Li, Zeming3; Zhang, Xiaoqin4; Weiming, Hu1,2,5 | |
2023-07 | |
会议名称 | International Conference on Machine Learning |
会议日期 | 2023-7 |
会议地点 | Honolulu, Hawaii, USA |
摘要 | Self-supervised learning on large-scale Vision Transformers (ViTs) as pre-training methods has achieved promising downstream performance. Yet, how much these pre-training paradigms promote lightweight ViTs' performance is considerably less studied. In this work, we develop and benchmark several self-supervised pre-training methods on image classification tasks and some downstream dense prediction tasks. We surprisingly find that if proper pre-training is adopted, even vanilla lightweight ViTs show comparable performance to previous SOTA networks with delicate architecture design. It breaks the recently popular conception that vanilla ViTs are not suitable for vision tasks in lightweight regimes. We also point out some defects of such pre-training, e.g., failing to benefit from large-scale pre-training data and showing inferior performance on data-insufficient downstream tasks. Furthermore, we analyze and clearly show the effect of such pre-training by analyzing the properties of the layer representation and attention maps for related models. Finally, based on the above analyses, a distillation strategy during pre-training is developed, which leads to further downstream performance improvement for MAE-based pre-training. Code is available at https://github.com/wangsr126/mae-lite. |
关键词 | Vision Transformer Self-supervised Learning Lightweight Networks Knowledge Distillation |
收录类别 | EI |
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/173211/52415 |
专题 | 多模态人工智能系统全国重点实验室_视频内容安全 |
通讯作者 | Gao, Jin |
作者单位 | 1.Institute of Automation, Chinese Academy of Sciences 2.School of Artificial Intelligence, University of Chinese Academy of Sciences 3.Megvii Research 4.Wenzhou University 5.ShanghaiTech University |
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 | Wang, Shaoru,Gao, Jin,Li, Zeming,et al. A Closer Look at Self-Supervised Lightweight Vision Transformers[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 | ||
ICML2023-MAE_Lite-ca(3478KB) | 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