May 18, 2022, by Lexi Earl
Swift: A ten minute coffee with Dr Michael Pound
Michael Pound was recently promoted to Associate Professor in the School of Computer Science. Mike was a Nottingham Research Fellow with the Future Food Beacon before becoming an Assistant Prof.
Tell us about your research in 10 words or less…
I use machine learning to quickly analyse biological images.
Now, explain your research for a lay audience
I work on the applied end of Computer Science, using machine and deep learning to quickly find and measure objects in images. Much of modern science is driven by image analysis, from microscopy up to large-scale aerial drone data. As a community we now produce more data than we can reasonably use, but it is important if we want to understand the biological mechanisms behind plants, and breed more resilient crops.
My research looks for faster and better ways to utilise modern machine learning practices to improve image capture and measurement. I often focus on newer deep learning techniques, in particular finding objects in images and segmentation regions of images. I specialise in high-resolution and high-dimensional data.
What do you enjoy most about your work?
There is always something new and exciting to work on! I have worked on segmentation of cells in microscope images, 3D segmentation of roots in soil, 3D reconstruction of leaves, analysis of hyperspectral images. I have always preferred broad research in computer vision than any specific small subset of it. Because the data I work on is very varied, the tools I use to solve these problems are also varied. This keeps it interesting, and there is always something new to learn.
What are the challenges?
Sometimes working in an applied area can be a challenge because you sit somewhat between two separate fields of research. It is easy to get lost between the two, I do not usually work on very specific problems in computer science, nor do I study the biology of plants directly. It can be a challenge to make sure that my research has useful impact in both fields, and not get side-tracked solving one or the other.
How have the challenges impacted on your approaches to your research questions?
It’s worked out ok so far! I often have to think carefully about the best way to solve a problem that also asks interesting questions in machine learning that will appeal to computer scientists. Sometimes it would be straightforward to use a standard algorithm for a task, but more interesting if we can work on a new way to improve beyond this. Often in biosciences and medicine there are challenges in these datasets that present opportunities to do something new and interesting, but you have to seek them out instead of falling back on the easy things.
Is there anything that you’ve found surprising in your work? Something you weren’t expecting perhaps?
The progress of deep learning and AI has been a constant surprise for years! Every time I think we have reached a plateau of performance, a new algorithm or technique appears that jolts everything forwards. Because my research is applied, I make use of many different techniques from across the field of computer vision, so there is always something new and surprising to try out.
What kind of impact has your research had/do you hope it will have?
I hope that my research has had a wide impact. Many of our approaches come with associated software tools to run them, but because I work in an interdisciplinary field, it isn’t always reasonable to assume someone running our research tools understands the underlying software and hardware requirements. We have always aimed to make it as easy to reproduce our work as possible, and to release our code as open source so people can adapt or extend it as they wish. Many of our tools and approaches have been downloaded thousands of times, which is really encouraging!
Do you work in collaboration with others? Who do you work with, and how does this impact the research?
All of the time! Since my work is interdisciplinary, the target of my research is often in another field, I’ve spoken about how I work with plant scientists, but I’ve also collaborated with engineers, physicists, psychologists, medical researchers, and industry. This is always varied and interesting work, but it means my research is often framed around a problem I don’t personally understand! My expertise is computer science, I often rely on my collaborators to bring knowledge of the problem domain. This is a very dynamic process, with regular meetings and discussions about progress, where we can and should go with the work, and how the results are helping or not.
Do you have any advice for future scholars?
Academia has days that are quiet, and days that are very busy, so remember to just take each job as it comes! Even on successful projects the research is never truly finished. There is usually more you could have done with more time, and that’s okay.
If you could change one thing about research, what would it be?
With the ongoing pandemic and concerns over climate change, research will need to look at the global way we collaborate, and try to find better ways to do this. Travelling to conferences is exciting and a great way to meet like-minded researchers, but we must include people who can’t travel, and so more study into the best use of online tools would be great.
No comments yet, fill out a comment to be the first
Leave a Reply