Self-Supervised Learning for Complex Visual Understanding

Program Overview

This PhD program focuses on advancing computer vision through self-supervised learning (SSL), a technique that enables models to learn from unlabeled data. The program aims to develop SSL methods for complex visual understanding tasks, such as autonomous navigation and medical image analysis. The successful candidate will have access to state-of-the-art facilities and work with a team of experienced researchers in AI and machine learning.

Program Outline

Other:

• This PhD project aims to advance the state-of-the-art in computer vision through the development and application of self-supervised learning (SSL) techniques.
• The focus will be on achieving complex visual understanding tasks without relying on extensive labelled datasets.
• Traditional supervised learning requires human experts to provide labelled examples and explicit instructions to train AI models.
However, obtaining large labelled datasets can be expensive and time-consuming. Most importantly, SL seriously restricts the machine's capability to generalise and learn new unseen scenarios.
• Inspired by human learning, SSL leverages the intrinsic structure within unlabelled data to train models effectively.
• Self-supervised learning is a machine learning paradigm where a model learns from the data without explicit external labels, creating a surrogate task that the model can learn to solve without the need for external annotations.
• Example Research Components:
• Contrastive Learning: The model is trained to differentiate between positive pairs (similar samples) and negative pairs (dissimilar samples).
• Representation Learning: learn rich and meaningful representations from visual data without explicit supervision.
The goal is to enable the model to automatically discover relevant features and hierarchical representations.
• Predictive Learning: The model is trained to predict certain parts of the input data.
e.g. predicting the next word in a sentence, filling in missing parts of an image.
• Generative Models: Training a model to generate a part of the input data or the entire input from some part of it, e.g.
generative adversarial networks (GANs). This can involve developing cross-modal self-supervised learning techniques.
• Temporal Learning: Learning representations based on the temporal order of data
• Transfer Learning and Generalization: Evaluate the transferability of learned representations across different domains and tasks.
• Real-world Applications and contribution: Apply the developed self-supervised learning techniques to real-world problems, such as autonomous navigation, scene understanding, or medical image analysis.
The project will evaluate the approach and model/algorithm performance on benchmark datasets and its application potential in collaboration with industry partners.
• Loughborough University and Department of Computer Science: Loughborough University has been awarded gold for student experience, gold for student outcomes and gold overall in the 2023 Teaching Excellence Framework- Triple Gold in TEF 2023.
The Department of Computer Science has an excellent research record in AI, machine learning, robotics, computer vision and data science.
• The successful candidate will have access to robotics and AI laboratories, high-spec computing facilities (e.g., A100 GPUs), HPC, and £5.8M DigLabs, complementing a £9m investment in research and teaching.
• You will have regular supervision meetings and work with a strong AI research team including over 30 PhDs/PDRAs/academic staff in the department.
• You will take part in various outreach and impact generation activities, and develop your career profile throughout your PhD study.

UK fee: £4,712 Full-time degree per annum International fee: £26,000 Full-time degree per annum University fees and charges can be paid in advance and there are several methods of payment, including online payments and payment by instalment. Fees are reviewed annually and are likely to increase to take into account inflationary pressures.