Dr Wanli Ma

About

I am a Postdoctoral Research Fellow at the University of Cambridge. I currently work within the Cambridge Centre for Smart Infrastructure and Construction (CSIC), under the supervision of Prof Brian Sheil, developing computer vision techniques (spanning image and video understanding to 3D/4D model generation) for measuring and monitoring construction productivity on site. My work bridges advances in visual reconstruction and scene understanding with the practical challenge of tracking construction progress, workforce activity, and site efficiency at scale.

Previously, at Cambridge I worked on AI-enabled large-scale disaster damage assessment (e.g., blasts, floods, and fires) as part of the EPSRC-funded MicroBlast project. I completed my PhD in Computer Science at Cardiff University in 2024, within the Visual Computing Group, under the supervision of Dr Oktay Karakus (Remote Sensing) and Prof Paul Rosin (Computer Vision, Fellow of the IAPR). My doctoral thesis, “Towards Minimal Supervision for Semantic Segmentation of Remote Sensing Imagery,” explored techniques to reduce the labelling burden in training machine learning models for satellite remote sensing imagery. Prior to my PhD, I received a Master’s degree with Distinction in Image and Video Communication and Signal Processing from the University of Bristol (2021), where my dissertation focused on “Ship Wake Detection in SAR Imagery Using Dual-Tree Complex Wavelet Transform (DT-CWT).”

My current research centres on computer vision for 3D/4D scene reconstruction — recovering spatial and temporal structure from images and video — and its application to construction productivity measurement, including site progress monitoring, activity recognition, and resource utilisation tracking. I am interested in bridging theoretical advances in AI with practical, real-world deployment, from earth observation to the built environment. My earlier work in remote sensing centred on “minimal supervision” approaches (semi-supervised learning, active learning, and multi-modal fusion) for large-scale disaster damage assessment, marine monitoring, and urban infrastructure detection — reducing the labelling burden in training machine learning models for satellite imagery. Beyond this, I retain a strong interest in applying computer vision to robotics and industrial automation, having previously developed and commercialised automated facilities for the manufacturing sector.

Skills

Recent News & Activities

Publications

Google Scholar

Oral & Poster Presentations

• Oral Presentation: The International Geoscience and Remote Sensing Symposium (IGARSS), Brisbane, Australia, 03-08 Aug 2025
• Oral Presentation: The International Geoscience and Remote Sensing Symposium (IGARSS), Athens, Greece, 07-12 Jul. 2024
• Oral Presentation: The British Machine Vision Conference (BMVC), Aberdeen, UK, 20-24 Nov. 2023
• Oral Presentation: The International Geoscience and Remote Sensing Symposium (IGARSS), Pasadena, US, 16-21 Jul. 2023
• Poster: The British Machine Vision Association (BMVA) Summer School, Norwich, UK, 11-15 Jul. 2022
• Poster: Vision Researchers Colloquium 2022, Cardiff, UK, 4 Jul. 2022
• Poster: Special Event: The British Machine Vision Association (BMVA) 3 Day Symposium, Manchester, UK, 4-6 Apr. 2022
• Oral Presentation: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Toronto, CA (Virtually), 6-11 Jun. 2021

Teaching and Supervision

Teaching

I have over two years of teaching experience in the computer science discipline, including
• Postgraduate module related to Machine Learning, enrollment: 150+ students(Since 2022)
• Undergraduate module related to Data Structures, enrollment: 150+ students (Since 2024)
• Summer courses “Python Programming” and “Machine Learning and Its Applications” (Since 2025)
  • Teaching two courses, each enrolling over 30 undergraduate students annually.
  • Python Programming: data type & structures, function, file I/O.
  • Machine Learning and Its Applications: focus on practical and theoretical aspects of machine learning, including supervised/unsupervised learning, neural networks, and model evaluation.
  • Machine Learning and Its Applications: guided students through a series of hands-on projects applying AI techniques to detect and analyse urban buildings from remote sensing imagery.

Supervision

• Diya Thomas, postgraduate at the University of Cambridge. Vision Foundation Models for Disaster Damage Assessment. (2026)
• Ningxin He, undergraduate at the National University of Singapore. Machine Learning Approaches for Earthquake-induced Building Damage Assessment. (2025)
• Xiaoyu Zhang, undergraduate at the University of Cambridge. Reinforcement Learning-based Control for Robotic Hoisting in Construction Environments (2025)
• Henry Booth, postgraduate at Cardiff University. Marine Debris Detection Using Deep Learning Networks. Now at the Met Office. (2023)
• Naga Padala, postgraduate at Cardiff University. Integrating PCA, HSV, and Raw Remote Sensing Data via Deep Learning for Land Cover Classification. (2022)

Collaboration

• University of Sheffield (2025 – present): Collaborating on blast-loading-informed building damage assessment, with the Sheffield team providing blast-loading simulations for the 2020 Beirut explosion area.
• Chinese Academy of Sciences (2025 - present): Co-developed landslide detection methods based on computer vision and satellite remote sensing imagery.
• Cardiff University (2025 - present):
  • Knowledge distillation from various foundation models for semantic segmentation of remote sensing imagery, with regular meetings to discuss technological developments.
  • Robot operation and hoisting: collaborating with Cardiff, which provides a simulation environment for soft-object manipulation, while we develop object state prediction and reinforcement learning–based control algorithms.
• AI for Good Lab – Microsoft Research (2023-2024): A collaboration focused on geospatial analytics, applying AI to extract actionable insights and structured information from spatial data, including satellite and aerial imagery. Specifically, we explored automated methods for selecting the most informative coresets from large-scale datasets for land cover mapping to reduce the workload for labelling.
• The German Aerospace Centre (DLR) (2023-2026):
  • Collaborated on coreset selection, with DLR providing large-scale satellite datasets for label cover mapping.
  • Collaborated on AI for physics-informed building damage assessment, with DLR providing building damage data for the 2020 Beirut Explosion.

Academic Services

• Reviewer for Artificial Intelligence Journals: IEEE Transactions on Image Processing (T-IP), IEEE Transactions on Neural Networks and Learning Systems (T-NNLS), IEEE Transactions on Circuits and Systems for Video Technology (T-CSVT), Neurocomputing, Digital Signal Processing (DSP).
• Reviewer for Remote Sensing Journals: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (J-STARS), International Journal of Applied Earth Observation and Geoinformation (JAG). GIScience & Remote Sensing (GRS), Remote Sensing (RS), Scientific Reports (SR).
• Reviewer for Robotics Journals: IEEE Transactions on Automation Science and Engineering (T-ASE).
• Session Chair: "Remote Sensing Image Reconstruction, Processing, and Enhancement" in IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 2025.
• Organiser for Workshops: Machine Vision for Earth Observation, Environmental Monitoring, and Climate Change in BMVC 2026.