research Camera-independent depth from defocus Single-image depth estimation from defocus blur that keeps working when you change the camera. WACV 2024. Sufficient frame subsets for video understanding Explaining video classifiers by finding the smallest set of frames that drives their decision. Multimodal echocardiogram classification Fusing echocardiogram video with clinical records to detect Takotsubo cardiomyopathy. EgoEMS A high-fidelity multimodal egocentric dataset for cognitive assistance in emergency medical services. AAAI 2026. In-home care system for Alzheimer's A real-time sensing and alerting system supporting people with Alzheimer's disease and their caregivers. Noise-robust speech emotion recognition CNNs and attention that keep recognising emotion from speech when the audio gets noisy. Smart Health 2021. side projects Social navigation of a mobile robot Getting a robot through a crowd using trajectory prediction and social forces. Hiding emotions from voice assistants Adversarial perturbations that hide emotion from a voice assistant without breaking speech-to-text. Auto-ML with Auto-Keras Letting neural architecture search design a CNN and comparing it with a hand-built one.