In daily routines, humans, not only learn and apply knowledge for visual tasks but also have intrinsic abilities to transfer knowledge between related vision tasks. For example, if a new vision task is relevant to any previous learning, it is possible to transfer the learned knowledge for handling the new vision task. In developing new computer vision algorithms, it is desired to utilize these capabilities to make the algorithms adaptable. Generally, traditional computer vision methods do not adapt to a new task and have to learn the new task from the beginning. These methods do not consider that the two visual tasks may be related and the knowledge gained in one may be applied to learn the other one efficiently in lesser time. Domain adaptation for computer vision is the area of research, which attempts to mimic this human behavior by transferring the knowledge learned in one or more source domains and use it for learning the related visual processing task in the target domain. Recent advances in domain adaptation, particularly in cotraining, transfer learning, and online learning have benefited computer vision research significantly. For example, learning from high-resolution source domain images and transferring the knowledge to learning low-resolution target domain information. This special issue will focus on the recent advances in domain adaptation for different computer vision tasks.
Topics of interest include, but are not limited to:
Domain adaptation for machine Learning frameworks for learning deep representations
Domain adaptation for face detection/recognition and tracking
Domain adaptation for object detection/ recognition and tracking
Domain adaptation and hybrid models for real-time computer vision tasks
Domain adaptation for human pose detection/recognition and estimation
Domain adaptation for event/action detection and recognition
Domain adaptation for few-shot learning
Domain adaptation for deep neural network optimization