VCG researchers have recently published a number of papers on learning in computer vision with limited supervision. Our paper in T-PAMI proposes a method for active learning by exploiting contextual data. The ECCV-18 work presents a framework for localizing activities in videos using weak supervision during training and our CVPR-19 paper shows how weak supervision can be used for video moment retrieval using text descriptions. In the T-IP paper, we show that information theoretic typicality can be exploited for identifying a minimal subset for manual labeling, while our paper in T-CSVT explored dataset creation with active learning. Below are links to the papers.

• Context-Aware Query Selection for Active Learning in Event Recognition, IEEE Trans. on Pattern Analysis and Machine Intelligence (T-PAMI), 2019.
• Weakly Supervised Video Moment Retrieval from Text Queries, IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 2019.
• W-TALC: Weakly-supervised Temporal Activity Localization and Classification, European Conference on Computer Vision (ECCV), 2018.
• Construction of Diverse Image Datasets from Web Collections with Limited Labeling, IEEE Trans. on Circuits and Systems for Video Technology (T-CSVT), 2019.
• Exploiting Typicality for Selecting Informative and Anomalous Samples in Videos, IEEE Trans. on Image Processing (T-IP), 2019.