Hi!

My name is Soufiane Belharbi. I am a post-doc at LIVIA Lab., ÉTS, Montreal in collaboration with McCaffrey Lab. / GCRC McGill. I am working with Eric Granger, Ismail Ben Ayed, and Luke McCaffrey on training neural networks with weak supervision.

Research Interests

• Machine learning
• Neural networks
• Representation learning
• Weakly supervised learning
• Interpretable machine learning

News

• Oct. 2024: WACV 2025, A Realistic Protocol for Evaluation of Weakly Supervised Object Localization. [arXiv][Code]
• Oct. 2024: AI and Digital Health Symposium, ETS Montreal, Canada. Incorporating Affective Computing to Enhance Health Assessments and Interventions. [Posters]
• Sept. 2024: NeurIPS 2024, SR-CACO-2: A Dataset for Confocal Fluorescence Microscopy Image Super-Resolution. [arXiv][Code]
• Aug. 2024: ECCVw 2024, Textualized and Feature-based Models for Compound Multimodal Emotion Recognition in the Wild. [arXiv][Code text-based][Code feature-based]
• Jul. 2024: New patent with Ericsson (Pending), Method and System for Providing Labeled Images for Small Cell Site Selection. [Link]
• Apr. 2024: CVPRw 2024, Source-Free Domain Adaptation of Weakly-Supervised Object Localization Models for Histology. [arXiv][Code]
• Apr. 2024: CVPRw 2024, Joint Multimodal Transformer for Dimensional Emotional Recognition in the Wild. [arXiv][Code]
• Mar. 2024: FG 2024, Guided Interpretable Facial Expression Recognition via Spatial Action Unit Cues. [arXiv][Code]
• Mar. 2024: FG 2024, Distilling Privileged Multimodal Information for Expression Recognition using Optimal Transport. [arXiv][Code]
• Mar. 2024: FG 2024, Subject-Based Domain Adaptation for Facial Expression Recognition. [arXiv][Code]
• Oct. 2023: Image and Vision Computing. DiPS: Discriminative Pseudo-Label Sampling with Self-Supervised Transformers for Weakly Supervised Object Localization. [arXiv][Code]
• Jul. 2023: Summer School 2023-Jul.3-7-Concordia University-Montreal, Canada. Leveraging Artificial Intelligence to Optimise Behavioural Health Interventions and Assessments. [Website]
• Mar. 2023: New arXiv, CoLo-CAM: Class Activation Mapping for Object Co-Localization in Weakly-Labeled Unconstrained Videos. [arXiv][Code]
• Mar. 2023: MELBA, Deep Weakly-Supervised Learning Methods for Classification and Localization in Histology Images: A Survey. [Webpage][PDF][Code][MIDL2023 poster]
• Nov. 2022: WACVw 2023, Discriminative Sampling of Proposals in Self-Supervised Transformers for Weakly Supervised Object Localization. [arXiv][Code]
• Oct. 2022: WACV 2023, TCAM: Temporal Class Activation Maps for Object Localization in Weakly-Labeled Unconstrained Videos. [arXiv][Code]
• Sep. 2022: Montreal AI Symposium, Constrained Sampling for Class-Agnostic Weakly Supervised Object Localization. [arXiv][Code]
• Aug. 2022: ICPR tutorial, Deep Learning Models for Weakly-Supervised Object Localization and Segmentation. [Page][Slides]
• Feb. 2022: MIDL 2022, Negative Evidence Matters in Interpretable Histology Image Classification. [arXiv][Code]
• Jan. 2022: WACV 2022, F-CAM: Full Resolution Class Activation Maps via Guided Parametric Upscaling. [arXiv][Code]
• Jan. 2022: TMI, Deep Interpretable Classification and Weakly-Supervised Segmentation of Histology Images via Max-Min Uncertainty. [arXiv][Code]

Phd

I completed my PhD in computer science at the Institut National des Sciences Appliquées de Rouen Normandie (INSA Rouen Normandie) in LITIS lab, Apprentissage (Learning) team (Oct. 2014-Jul.2018). I was supervised by Prof. Sébastien Adam (director), Clément Chatelain, and Romain Hérault. During my PhD, I conducted research on the regularization of neural networks through representation learning with particular focus on learning scenarios where only few training samples are available. [BibTeX] [ArXiv][HAL-Thèse][Presentation]

@phdthesis{sbelharbiphd2018,
  author = {Belharbi, S.},
  title = {Neural Networks Regularization Through Representation Learning},
  school = {Normandie Université, INSA Rouen Normandie, LITIS laboratory},
  year = {2018},
  pages = {196},
  note = {Supervisor: Sébastien Adam, Advisors: Clément Chatelain, Romain Hérault},
  url = {https://arxiv.org/pdf/1807.05292}
}