Jan van Gemert

Ternary multiplicative adaptation recovers lost performance in quantized models while maintaining extreme efficiency, outperforming traditional low-bit methods.

Motion cues derived from event camera features can dramatically enhance optical flow accuracy, especially in data-scarce environments.

EventKitchen reveals the complexities of real-world cooking activities, setting a new benchmark for event-based perception that challenges existing datasets focused on scripted actions.

HAR models can exhibit statistically significant biases based on skin color, even when performing the exact same action.

Autoencoders can reconstruct rare image features more faithfully by explicitly focusing on statistically uncommon spatial locations during training.