On 14 May 2026, Professor Vasileios Belagiannis gave a keynote talk on ‘Memory and Uncertainty Quantification in the Optimisation of Generative Models’ at the 6th Workshop on Advanced Imaging and Visualisation (CAMPing 2026) in Faak am See, Austria. The presentation showed the group’s recent advances in improving the reliability and efficiency of deep learning models. He specifically discussed the group’s recent methodology for gradient-based uncertainty estimation in depth estimation, as well as post-hoc distribution shift detection for robust trajectory forecasting in automated driving. He also presented the group’s work on neural architecture search, which uses discrete graph diffusion to generate high-performing and hardware-aware architectures. Furthermore, the talk addressed the issue of memorisation in diffusion models by proposing a denoising-free detection metric based on log-probability anisotropy. Finally, he demonstrated how pixel-wise aleatoric uncertainty estimation can optimise the generative sampling process to improve image quality.
Related Publications
* R. Asthana & V. Belagiannis. “Detecting and Mitigating Memorization in Diffusion Models through Anisotropy of the Log-Probability.” *International Conference on Learning Representations (ICLR)*, 2026.
* M. De Vita, J. Wiederer, & V. Belagiannis. “Forecasting the Past: Gradient-Based Distribution Shift Detection in Trajectory Prediction.” *CVPR Workshops*, 2026.
* J. Hornauer, A. El-Ghoussani, & V. Belagiannis. “Revisiting Gradient-Based Uncertainty for Monocular Depth Estimation.” *IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)*, 2025.
* R. Asthana, J. Conrad, M. Ortmanns, & V. Belagiannis. “Dextr: Zero-Shot Neural Architecture Search with Singular Value Decomposition and Extrinsic Curvature.” *Transactions on Machine Learning Research (TMLR)*, 2025.
* M. De Vita & V. Belagiannis. “Diffusion Model Guided Sampling with Pixel-Wise Aleatoric Uncertainty Estimation.” *IEEE Winter Conference on Applications of Computer Vision (WACV)*, 2025.
* Y. Yao, S. Bhatnagar, M. Mazzola, V. Belagiannis, I. Gilitschenski, L. Palmieri, S. Razniewski, & M. Hallgarten. “AGENTS-LLM: Augmentative Generation of Challenging Traffic Scenarios with an Agentic LLM Framework.” *IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)*, 2025.
* M. Hölle, W. Kellermann, & V. Belagiannis. “Uncertainty-Aware Likelihood Ratio Estimation for Pixel-Wise Out-of-Distribution Detection.” *ICCV Workshops*, 2025.
* R. Asthana, J. Conrad, Y. Dawoud, M. Ortmanns, & V. Belagiannis. “Multi-conditioned Graph Diffusion for Neural Architecture Search.” *Transactions on Machine Learning Research (TMLR)*, 2024.
* J. Hornauer & V. Belagiannis. “Heatmap-based out-of-distribution detection.” *IEEE Winter Conference on Applications of Computer Vision (WACV)*, 2023.
* J. Hornauer & V. Belagiannis. “Gradient-based Uncertainty for Monocular Depth Estimation.” *European Conference on Computer Vision (ECCV)*, 2022.
* J. Wiederer, A. Bouazizi, M. Troina, U. Kressel, & V. Belagiannis. “Anomaly detection in multi-agent trajectories for automated driving.” *Conference on Robot Learning (CoRL)*, 2022.