Self-supervised and cross-modal learning from videos

Thesis

Deep learning has demonstrated impressive results for tasks where the training of neural networks can be supervised with paired input and manually labelled output data. However, labelling data can be expensive and might not be feasible for some applications. In this thesis, we consider learning from video data using less supervision than standard supervised learning methods. In particular, we focus on self-supervised learning, where the data itself provides supervision without requiring ma...

Expand abstract

Authors: Koepke, AS

• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: cross-modal learning; self-supervised learning
• Subjects: Computer vision; Machine learning