Selected talks, conference presentations, posters, project updates, and research seminars.
Workshop Presentation, QualITA Workshop | ICPE 2026 , Florence, Italy
This talk presents ShapBPT in Perspective, a consolidated review and practical case study of ShapBPT for eXplainable Anomaly Detection (XAD).
Conference Talk, Singapore Expo Centre , Singapore
Introduces ShapBPT, a data-aware hierarchical Shapley-based method for image feature attribution that improves efficiency, structural alignment, and interpretability in computer vision models.
Consortium Meeting, Universidad de Vigo , Vigo, Spain
A consortium presentation on VAE-GAN based unexpected condition detection for industrial safety monitoring in robotic environments, developed within UC3 of the DistriMuSe EU project.
Conference Talk, Mediterranean Conference Centre , La Valletta, Malta
A case study on the trustworthiness of VAE-based anomaly detection systems using Explainable AI, highlighting how high anomaly scores can be driven by misleading visual features.
Talk, Vancouver Convention Centre , Vancouver, Canada
We investigate the use of a stratified sampling approach for LIME Image, a popular model-agnostic explainable AI method for computer vision tasks, in order to reduce the artifacts generated by typical Monte Carlo sampling. Such artifacts are due to the undersampling of the dependent variable in the synthetic neighborhood around the image being explained, which may result in inadequate explanations due to the impossibility of fitting a linear regressor on the sampled data. We then highlight a connection with the Shapley theory, where similar arguments about undersampling and sample relevance were suggested in the past. We derive all the formulas and adjustment factors required for an unbiased stratified sampling estimator. Experiments show the efficacy of the proposed approach..