About Me
I am a PhD researcher at IDLab (University of Antwerp) working at the intersection of hyperspectral imaging and deep learning. My research focuses on data-efficient representation learning and model interpretability so that hyperspectral methods remain reliable outside the lab. In particular, my thesis spans (i) multi-label patch-level classification, (ii) self-supervised/contrastive learning under scarce labels, and (iii) explainability-driven band selection for dimensionality reduction.
I enjoy turning principled ideas into practical tools that work across domains. Much of my work focuses on pre-training with minimal or no annotation, followed by fine-tuning robust predictors, and utilising explanation signals to identify spectrally meaningful bands while maintaining performance. I’m especially interested in evaluation that goes beyond headline accuracy—thinking about stability, uncertainty, and whether explanations align with domain knowledge.
Before moving into AI, I completed an MSc in Money & Banking (American University of Beirut) and worked in international banking. As a CFA charterholder, I bring rigour, risk thinking, and an eye for operational constraints to model design and validation. That background has shaped how I prioritise reproducibility, clarity, and decision-usefulness in research.
Recent outputs include a journal article on training strategies for multi-label hyperspectral classifiers and work on contrastive learning in low-label regimes and explainability-guided band selection. I also review for IEEE TGRS and have supervised BSc and MSc projects at the University of Antwerp.
In my free time, I enjoy exploring the latest advancements in AI and machine learning, as well as engaging in discussions about their ethical implications.
Research Interests
- Artificial Intelligence
- Machine Learning
- Hyperspectral Image Analysis
- Computer Vision
- Explainable and Reliable ML
- Data Science
Hobbies
- Reading
- Photography
- Walking
- Hiking
- Travelling
- Bread baking
Publications
-
Training Methods of Multi-label Prediction Classifiers for Hyperspectral Remote Sensing Images
Haidar Salma, Oramas Mogrovejo José Antonio
Remote sensing - ISSN 2072-4292 - 15:24(2023), 5656 -
A Contrastive Learning Method for Multi-Label Predictors on Hyperspectral Images
Haidar Salma, Oramas Mogrovejo José Antonio
2023 13th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing (WHISPERS), 31 October 2023 - 02 November 2023, Athens, Greece - ISSN 2158-6276 - IEEE, 2023, p. 1-5 -
Enhancing Hyperspectral Image Prediction with Contrastive Learning in Low-Label Regime
Salma Haidar, José Oramas
arXiv, 2023 -
Explainability-Driven Dimensionality Reduction for Hyperspectral Imaging
Salma Haidar, José Oramas
arXiv, 2025
Scientific Services and Activities
- Volunteer reviewer for IEEE Transactions on Geoscience and Remote Sensing (TGRS).
- Track Lead (Data & ML), CloudBrew Conference (Mechelen, Belgium): curate the technical programme, review proposals, and moderate sessions.
Contact
Email: salma.haidar@uantwerpen.be
University of Antwerp, imec-IDLab
Sint-Pietersvliet 7
Antwerp, B2000
Belgium.