UCL Department of Civil, Environmental and Geomatic Engineering


Geomatics researchers win Best Paper Award

27 July 2021

PhD student David Griffiths and Dr Jan Boehm received the Remote Sensing MDPI 2021 Best Paper Award

A photo of David Griffiths

We are delighted to announce that Dr Jan Boehm and David Griffiths won the Remote Sensing MDPI 2021 Best Paper Award. The paper, “A Review on Deep Learning Techniques for 3D Sensed Data Classification” has been downloaded over 15,000 times. We caught up with David to find out more.  

  • What was your paper on? 

Our paper is called, “A Review on Deep Learning Techniques for 3D Sensed Data Classification”. It covers the recent boom in deep learning approaches for 3D data processing.  

Deep learning is a machine learning technique which is responsible for the rapid growth of Artificial Intelligence in the past decade. It provides a way for machines to learn to solve very complicated problems (in particular, ones humans are unable to teach a computer to solve), by providing it pairs of input (e.g., an image) and the desired output (e.g., where a person is in the image). It has witnessed very large success in 2D image problems such as facial recognition, automatic image retrieval / searching and detecting objects present in the image.  

There have been great advancements using this technique for 3D data as well. Examples of 3D problems range from autonomous driving, robotic hovers to virtual and augmented reality, to name a few. Deep learning with 3D data will allow computers to understand what is around them, and build a map of this, allowing it to navigate around and solve any number of tasks. 

The paper specifically looked at the recent advancements from classical machine learning models to deep-learning based models. We outlined the full spectrum of research direction undertaken in the field to achieve this, whilst also offering a detailed discussion of future research directions. Our paper was one of the first review papers dedicated specifically to deep learning and 3d data classification. 

  • What do you think the impact of your research is/will be?  

The paper can act as first port of call for students and new researchers entering the field of 3D data processing using modern deep-learning methods. This is a rapidly growing field and as such, our paper has already received over 15,000 downloads and been cited by the academic community over 70 times. As it covers a transition period from more traditional methods to modern approaches, we expect the paper to stay relevant for a long time to come. 

  • What have you enjoyed the most about doing your PhD? 

The thing I enjoy most about my PhD is the freedom of research it enables. I get to spend my whole day researching and building bleeding-edge technologies. Having no strict research agenda or deadlines creates the perfect environment to be creative and curious. This is largely a result of having such an excellent supervisor. 

  • What have you found the most challenging about doing your PhD?  

The most challenging aspect of my work is the long hours I inevitably spend working on a computer. Being someone who would naturally always choose to be outside rather than inside, I sometimes find the long days very draining. Fortunately, regular running around Hampstead Heath offers some relief!