O Abstract:
Deep learning has made remarkable advancements in various computer vision tasks, such as object recognition, detection, segmentation, and person re-identification. However, the performance of these models can deteriorate when they are applied to new domains with different distributions or characteristics than the data they were trained on. This is known as the domain shift problem and can pose a significant challenge in real-world scenarios, particularly in defense and security applications. In this talk, I will briefly present my research on addressing the domain shift problem through feature disentanglement, meta-learning, image augmentation, and regularization techniques.
