Image: Khadija Iddrisu
PhD candidate Khadija Iddrisu discusses the challenge of ‘motion blur’ in computer-vision research and how the tech benefits the sports, medicine and driving sectors.
Khadija Iddrisu is in the early stages of a PhD at the Insight SFI Research Centre for Data Analytics in Dublin City University (DCU). Having completed an undergraduate degree in computer science, she undertook a master’s degree at the African Institute for Mathematical Sciences (AIMS) where she applied computer-vision techniques to the problem of brain disease diagnosis.
The success of this project opened up opportunities for Iddrisu to further develop her knowledge and skills “in preparation for a long-term research career”, as she puts it.
‘Given the novelty of my research field, finding publicly available datasets can be challenging’
Tell us about your current research.
In the past, my work primarily focused on the applications of machine learning in medical imaging, specifically brain scans. Upon my arrival at DCU, I had the opportunity to visit Xperi, my PhD industry collaborator, and gain insight into the diverse range of research undertaken there.
A notable revelation was my introduction to event-based vision, a concept and technology I was unfamiliar with prior to the visit.
Witnessing the operation of these innovative cameras, which seemed revolutionary in terms of their applicability in object detection due to their numerous properties, piqued my interest immediately.
To gain experience using this data with deep learning, I have worked on replicating Xperi research that utilised event cameras for face-tracking and blink-detection. Through this replication process, I have learned to navigate the challenge of limited publicly available event datasets using event simulators and understood the complexities of applying deep-learning models, traditionally used to 2D frames, to event streams.
Additionally, I have conducted extensive literature research to explore the wide range of applications of event cameras in machine learning and computer vision.
These experiences have allowed me to refine my research interests for my PhD journey. I am currently in the process of defining a focused PhD topic and outlining the research questions that will guide the next stages of my four-year PhD journey.
In your opinion, why is your research important?
My research bears significant importance due to its potential direct applicability in addressing societal issues.
One prevalent challenge in computer vision and frame-based cameras is ‘motion blur’. This phenomenon occurs when information is lost due to the rapid movement of an object, seen in scenarios involving spinning, driving or sudden, unexpected movements. Conventional frame-based cameras may miss the position of such objects due to their low frame rates, making image reconstruction in fast-moving objects a challenging task.
Event cameras, with their high dynamic range, high temporal resolution and low latency, resolve the issue of motion blur.
These properties open avenues for applications in high-speed tracking, fast image reconstruction and more.
I foresee the integration of this technology in driver-monitoring systems to mitigate road accidents and in autonomous vehicles to adeptly manipulate changing light scenes or sudden occlusions.
The sports sector can also leverage these cameras to gain detailed insights into players’ techniques and game strategies.
Moreover, medical procedures requiring adaptive lighting conditions or early disease detection through changes in skin colour can also benefit from this technology.
I am eager to embark on a project that leverages these innovative applications to address real-world problems, contributing to the betterment of the global community.
What inspired you to become a researcher?
When I was first introduced to artificial intelligence and machine learning, I was immediately intrigued and eager to start utilising it to solve problems. However, the direction in terms of choosing between a research or industry career was unclear. After careful consideration, I decided to pursue a path that would maximise the potential impact of my work.
Fortunately, toward the end of my master’s studies in mathematical sciences, I engaged in a research project that garnered considerable attention.
I worked on utilising deep learning approaches for brain vessel segmentation, a significant challenge in the process of diagnosing brain diseases.
I had the honour of representing my home country, Ghana, at a conference to present my research, receiving invitations to several events to discuss my work.
Although the research was still in progress, experiencing the direct impact of my work on people’s lives was incredibly fulfilling. This experience profoundly inspired me to delve deeper into research, having identified it as my tool for making an impact on people’s lives.
The exemplary work and mentorship of others in the field further motivated me, solidifying my resolve to pursue a career in research.
What are some of the biggest challenges or misconceptions you face as a researcher in your field?
One challenge prevalent in my research field is the scarcity of resources, such as datasets, computational power and substantial research publications on the topic.
Given the novelty of my research field, finding publicly available event datasets for experiments can be challenging, as is acquiring the skillset to manipulate available resources.
Fortunately, being part of the industry collaboration between Insight and Xperi provides me access to such resources, for which I am grateful.
Do you think public engagement with science has changed in recent years?
I believe public engagement with science has actually transformed significantly in recent years, in light of the Covid-19 pandemic.
In the past, most conferences were in person, limiting access to researchers in certain parts of the world. I missed out on opportunities due to challenges with travel funding, timing, etc.
However, with enhanced connectivity by digital and virtual conference platforms, there has been more access to scientific discourse, enabling a more informed and participative public.
Most researchers like me are now presented with the opportunity to share their work and gain engagement by remote participation in conferences, the use of social media sites such as YouTube, TikTok and LinkedIn to gain public engagement and reach to out directly to people who would benefit from their work.
To foster engagement with my own work, I am committed to prioritising accessibility and outreach. I actively utilise online platforms, mostly Twitter and LinkedIn, to disseminate my research insights, provide updates on my research journey to share with like-minded people the challenges and gains, and engage with diverse audiences, striving to present my findings in a concise and accessible manner.
I also utilise other platforms such as machine learning communities to train and mentor people who are considering a similar career path. This initiative allows me to make an impact while gaining understanding of challenges I can solve with my research ideas.
By adopting a multifaceted engagement approach, I aim to enrich public understanding, fuel curiosity and learning, and contribute to fostering a culture of informed and inclusive scientific dialogue.
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.
