Publications

The following is a list of my publications. I currently work on the application of representation learning methods to earth observation and computer vision with tools such as Optimal Transport. The complete list of publications can be found on my Google Scholar account.

Contributions to Representation Learning in Computer Vision and Remote Sensing

13/12/2024 · 1 minute · Paul Berg

I successfully defended my thesis.

Horospherical Learning with Smart Prototypes

25/11/2024 · 2 minutes · Paul Berg, Björn Michele, Minh-Tan Pham, Laetitia Chapel, Nicolas Courty

Published as an oral at BMVC 2024. Paper Code Hyperbolic spaces have emerged as an effective manifold to learn representations due to their ability to efficiently represent hierarchical data structures, with little distortion, even for low-dimensional embeddings. In the chosen hyperbolic model, such as the Poincaré ball, classification is usually conducted by leveraging a signed distance function to the hyperbolic equivalent of a plane (gyroplanes) or by measuring the alignment to a virtual fixed prototype.

Box for Mask and Mask for Box: weak losses for multi-task partially supervised learning

27/11/2024 · 1 minute · Hoang-Ân Lê, Paul Berg, Minh-Tan Pham

Published as a poster at BMVC 2024. Paper Code Object detection and semantic segmentation are both scene understanding tasks yet they differ in data structure and information level. Object detection requires box coordinates for object instances while semantic segmentation requires pixel-wise class labels. Making use of one task’s information to train the other would be beneficial for multi-task partially supervised learning where each training example is annotated only for a single task, having the potential to expand training sets with different-task datasets.

Multimodal Supervised Contrastive Learning in Remote Sensing Downstream Tasks

15/04/2024 · 1 minute · Paul Berg, Baki Uzun, Minh-Tan Pham, Nicolas Courty

Based on the previous article, we generalize our multi-modal contrastive learning framework to the supervised setting.

Joint multi-modal Self-Supervised pre-training in Remote Sensing: Application to Methane Source Classification

19/06/2023 · 1 minute · Paul Berg, Minh-Tan Pham, Nicolas Courty

In the task of multi-modal classification, we introduce a contrastive learning framework which is based on the regular contrastive learning from computer vision.

Spherical Sliced Wasserstein

01/05/2023 · Clément Bonet, Paul Berg, Nicolas Courty, Lucas Drumetz, François Septier, Minh-Tan Pham

Self-Supervised Learning for Scene Classification in Remote Sensing: Current State of the Art and Perspectives

17/08/2022 · 1 minute · Paul Berg, Minh-Tan Pham, Nicolas Courty

We review the current state of the art in self-supervised pre-training for computer vision and remote sensing. Peculiarities in remote sensing data often require the development of bespoke methods.

Weakly Supervised Detection of Marine Animals in High Resolution Aerial Images

12/01/2022 · 2 minutes · Paul Berg, Deise Santana Maia, Minh-Tan Pham, Sébastien Lefèvre

Paper Code Human activities in the sea, such as intensive fishing and exploitation of offshore wind farms, may impact negatively on the marine mega fauna. As an attempt to control such impacts, surveying, and tracking of marine animals are often performed on the sites where those activities take place. Nowadays, thank to high resolution cameras and to the development of machine learning techniques, tracking of wild animals can be performed remotely and the analysis of the acquired images can be automatized using state-of-the-art object detection models.