November 19, 2025 13:00 CEST
Live on Microsoft Teams.
On 19 November at 13:00 CEST, the ESA Φ-Lab Collaborative Innovation Network will host a new Φ-talk. Details are below.
Meet the speakers
Mélisande Teng is a PhD candidate in Computer Science at Université de Montréal / Mila - Quebec AI Institute and her research focuses on applications of machine learning for biodiversity monitoring. In particular, she has worked on species distribution modeling using remote sensing data, tree monitoring with drone imagery and bioacoustics. She aims at bridging the machine learning and ecology communities and the general public. She holds an MSc in Mathematics, Vision and Learning (MVA) from ENS Paris-Saclay, an MSc in Engineering from Ecole Centrale Paris and an Msc in Management and Social Entrepreneurship from ESSEC Business School.
Talk abstract
Information on trees at the individual level is crucial for monitoring forest ecosystems and planning forest management. Current monitoring methods involve ground measurements, requiring extensive cost, time and labour. Advances in drone remote sensing and computer vision offer great potential for mapping individual trees from aerial imagery at broad scale. Large pre-trained vision models, such as the Segment Anything Model (SAM), represent a particularly compelling choice given limited labelled data. In this talk, we present a study in which we investigate when and how SAM is useful for the task of tree crown instance segmentation in high-resolution drone imagery. We compare methods leveraging SAM in three use cases: 1) boreal plantations, 2) temperate forests, and 3) tropical forests. We also look into integrating elevation data into models, in the form of Digital Surface Model (DSM) information, which can readily be obtained at no additional cost from RGB drone imagery. We present BalSAM, a model leveraging SAM and DSM information, which shows potential over other methods, particularly in the context of plantations. We find that methods using SAM out-of-the-box do not outperform a custom Mask R-CNN, even with well-designed prompts. However, efficiently tuning SAM further and integrating DSM information are both promising avenues for tree crown instance segmentation models.