The scientific seminars of the COMET TSI are back for 2025-2026!
Please note that the seminars will now be hosted on Teams, and announced here: https://www.milo-seminars.org/. If you want to be systematically informed of the next seminars, you can simply send an email to join the mailing list of the seminars.
For the next seminar, we are pleased to welcome Antoine Saget (University of Strasbourg) on Tuesday 9 December from 13:30 to 13:40 by videoconference!
Title: Contrastive self-supervised learning on satellite image time series: application to crop classification with Sentinel-2 data
Abstract: The abundance of unlabeled Satellite Image Time Series (SITS) from missions such as Sentinel-2 contrasts with the scarcity of expert annotations, making self-supervised learning (SSL) an appealing strategy for large-scale remote sensing. Contrastive SSL is a natural way to exploit such data, yet its application to SITS remains challenging: data augmentation, a key component of contrastive learning, is less established for time series than for natural images. We first introduce FranceCrops, a large-scale dataset of 6.3 million agricultural parcels designed for SSL on SITS, providing unlabeled data for pre-training alongside labeled subsets for finetuning and evaluation. Then, we propose two methods to make contrastive SSL effective on SITS. First, GaPP (Groups as Positive Pairs) bypasses augmentations altogether by exploiting pre-existing groupings within the data to form positive pairs. Second, a temporal resampling augmentation creates two distinct yet coherent views of the same time series. Together, these approaches reach state-of-the-art performance in low-label settings on crop classification tasks and highlight the importance of positive-pair quality in SITS representation learning.
Videoconference link: MILO seminar - Contrastive SSL - Antoine Saget | Réunion-Joindre | Microsoft Teams
Looking forward to seeing you on Tuesday 9!
Pour participer à cet événement, l’inscription est obligatoire.
Partager :
