规模一致和临时组合无监督域调整用于对象检测
Lunfeng Guo1,2,3, Yizhe Zhang1,2,3, Jiayin Liu1,4
1School of Mechanical and Electrical Engineering, China University of Mining and Technology (Beijing), Beijing 100083, China.
Sensors (Basel, Switzerland)
|January 11, 2025
概括
本研究引入了一个新的无监督域适应对象检测 (UDA-OD) 框架,该框架可以提高小物体检测和稳定性. 这种新的方法使用了尺度一致性和时间伪标签选择,以更好地跨领域适应.
科学领域:
- 计算机视觉 计算机视觉
- 机器学习 机器学习
- 人工智能的人工智能
背景情况:
- 无监督域适应对象检测 (UDA-OD) 解决域移动,但与小对象和不准确的伪标签作斗争.
- 现有的UDA-OD方法往往过度依赖于分类信心,导致局限框定位差.
研究的目的:
- 开发一个新的UDA-OD框架,增强跨领域的稳定性和检测性能,特别是对于小物体.
- 为了提高伪标签选择和界限框定位在UDA-OD中的准确性.
主要方法:
- 引入跨尺度预测一致性 (CSPC) 以在多个分辨率上进行强大的检测.
- 集成类内部特征一致性 (ICFC) 使用对比学习来对齐特征表示.
- 开发了时间合并伪标签选择 (TEPLS),将时间稳定性和分类信心结合起来,以获得高质量的伪标签.
主要成果:
- 在具有挑战性的UDA-OD基准 (Cityscapes,Sim10k,虚拟矿) 上取得了最先进的表现.
- 在小物体检测准确度方面取得了显著的改进.
- 与现有方法相比,展示了增强的跨领域整体稳定性.
结论:
- 拟议的UDA-OD框架有效地解决了当前方法的局限性.
- 尺度一致性和高级伪标签选择的结合显著提高了检测性能.
- 该方法为现实世界对象检测挑战提供了强大的解决方案,其中包括域移动.
相关概念视频
01:00Detection of Gross Error: The Q Test
When one or more data points appear far from the rest of the data, there is a need to determine whether they are outliers and whether they should be eliminated from the data set to ensure an accurate representation of the measured value. In many cases, outliers arise from gross errors (or human errors) and do not accurately reflect the underlying phenomenon. In some cases, however, these apparent outliers reflect true phenomenological differences. In these cases, we can use statistical methods...
01:16Combining Functions
Functions can be combined to form new mathematical models that describe interactions between variables. These combinations are fundamental in understanding relationships between changing quantities and are commonly encountered in scientific and engineering contexts. The combination methods—addition, subtraction, multiplication, division, and composition—each have unique implications for the resulting function’s domain and behavior.When combining functions through arithmetic operations, such...