Variational Information Bottleneck for Cross Domain Object Detection
- Resource Type
- Conference
- Authors
- Chen, Jiangming; Deng, Wanxia; Peng, Bo; Liu, Tianpeng; Wei, Yingmei; Liu, Li
- Source
- 2023 IEEE International Conference on Multimedia and Expo (ICME) ICME Multimedia and Expo (ICME), 2023 IEEE International Conference on. :2231-2236 Jul, 2023
- Subject
- Communication, Networking and Broadcast Technologies
Components, Circuits, Devices and Systems
Computing and Processing
Signal Processing and Analysis
Costs
Filtering
Object detection
Detectors
Data mining
Task analysis
Information theory
Cross Domain Object Detection
Unsupervised Domain Adaptation
Variational Information Bottleneck
Feature Disentanglement
- Language
- ISSN
- 1945-788X
Cross domain object detection leverages a labeled source domain to learn an object detector which performs well in a novel unlabeled target domain. Most existing works mainly align the distribution utilizing the entire image knowledge ignoring the obstacles of task-uncorrelated information to alleviate the domain discrepancy. To tackle this issue, we propose a novel module called Variational Instance Disentanglement (VID) based on information theory which aims to decouple the information of task-correlated while filtering out the task-uncorrelated factors at the instance level. Notably, the proposed VID can be used as a plug-and-play module without bringing extra network parameter cost. We equip it with adversarial network and self-training network forming Variational Instance Disentanglement Adversarial Network (VIDAN) and Variational Instance Disentanglement Self-training Network (VIDSN), respectively. Extensive experiments on multiple widely-used scenarios show that the proposed method improves the performance of the popular frameworks and outperforms state-of-the-art methods.