Benjamin Alexander Kellenberger
Domaines de compétences
- Earth observation
- Wildlife detection, monitoring, and conservation (from drones, camera traps, and more)
- Machine learning, deep learning; Convolutional Neural Networks
- Software engineering; interfaces
- 2020–: Postdoctoral Researcher, EPFL, Sion, Switzerland
- 2020: Postdoctoral researcher, Wageningen University, Netherlands
- 2017–2020: PhD (cont'd); Wageningen University, Netherlands
- 2016–2017: PhD; University of Zurich, Switzerland
- 2009–2014: BSc. and MSc. in geography (remote sensing and GIS) and computer science; University of Zurich, Switzerland
Sélection de publications
|Tuia, D.*, Kellenberger, B.*, Beery, S.*, Costelloe, BR*, Zuffi, S., Risse, B., Mathis, A., Mathis, MW, van Langevelde, F., Burghardt, T., Kays, R., Klinck, H., Wikelski, M., Couzin, ID, van Horn, G., Crofoot, MC, Stewart, CV, Berger-Wolf, T.
* equal contribution
Nature Communications, 2022
|Perspectives in Machine Learning for Wildlife Conservation|
|Kellenberger, B., Tuia, D., Morris, D.
Methods in Ecology and Evolution, 2020
|AIDE: Accelerating image-based ecological surveys with interactive machine learning.|
|Kellenberger, B., Marcos, D., Lobry, S., Tuia, D.
IEEE Transactions on Geoscience and Remote Sensing, 2019
|Half a Percent of Labels is Enough: Efficient Animal Detection in UAV Imagery using Deep CNNs and Active Learning.|
|Kellenberger, B., Marcos, D., Tuia, D.
IEEE Conference on Computer Vision and Pattern Recognition workshops, 2019
|When a Few Clicks Make All the Difference: Improving Weakly-Supervised Wildlife Detection in UAV Images.|
|Kellenberger, B., Marcos, D., Tuia, D.
Remote Sensing of Environment, 2018
|Detecting mammals in UAV images: Best practices to address a substantially imbalanced dataset with deep learning.|
|Damodaran, BB, Kellenberger, B., Flamary, R., Tuia, D. (joint first author)
European Conference on Computer Vision, 2018
|DeepJDOT: Deep Joint Distribution Optimal Transport for Unsupervised Domain Adaptation.|
|Marcos, D., Tuia, D., Kellenberger, B., Zhang, L., Bai, M., Liao, R., Urtasun, R.
IEEE Conference on Computer Vision and Pattern Recognition, 2018
|Learning Deep Structured Active Contour Models End-to-end.|
|Marcos, D., Volpi, M., Kellenberger, B., Tuia, D.
ISPRS Journal of Photogrammetry and Remote Sensing, 2018
|Land Cover Mapping at Very High Resolution with Rotation Equivariant CNNs: Towards Small yet Accurate Models.|
AIDEAIDE is a web-based, collaborative image annotation platform with tight and seamless integration of machine (deep) learning models through active learning.
- AIDE supports annotating billions of images with labels, points, bounding boxes, or pixel-wise segmentation masks
- Concurrency: an arbitrary number of users can annotate images at the same time
- User-provided image annotations are automatically used to train a deep learning model in the background. This model then predicts the pool of unlabelled images and returns those to the users for annotating that are most interesting (active learning)
- Popular models (ResNet, RetinaNet, U-Net) are built-in and require zero coding experience to use; custom models can be added with minimal overhead
- Models and data can be managed, trained, evaluated in accuracy, and shared, all through the web browser
You can get it here: GitHub
Paper (please cite when using AIDE): PDF
- Digital wildlife conservation: I like to know what users of my systems need for actual conservation, which may involve more than just detection models.
- Habitats in a changing world: Beyond conservation, I am interested in ﬁnding out more about how species interact with their environment and themselves, and how their distribution may be affected due to changes. Remote sensing may play a vital role in this quest.
- Humans in the loop: machine learning is great, but inoperable without signiﬁcant input by humans. I believe their role in the process of "automating everything" has been neglected for too long…