Unsupervised Domain Adaptation for Training Event-Based Networks Using Contrastive Learning and Uncorrelated Conditioning
Dayuan Jian, Mohammad Rostami
Abstract
Event-based cameras offer reliable measurements for preforming computer vision tasks in high-dynamic range environments and during fast motion maneuvers. However, adopting deep learning in event-based vision faces the challenge of annotated data scarcity due to recency of event cameras. Transferring the knowledge that can be obtained from conventional camera annotated data offers a practical solution to this challenge. We develop an unsupervised domain adaptation algorithm for training a deep network for event-based data image classification using contrastive learning and uncorrelated conditioning of data. Our solution outperforms the existing algorithms for this purpose.
Topics & Concepts
Computer scienceUncorrelatedArtificial intelligenceEvent (particle physics)Domain adaptationAdaptation (eye)Domain (mathematical analysis)Deep learningMachine learningComputer visionPattern recognition (psychology)StatisticsMathematicsPhysicsOpticsMathematical analysisClassifier (UML)Quantum mechanicsAdvanced Memory and Neural ComputingAdvanced Neural Network ApplicationsMachine Learning and ELM