Publications by authors named "Xinxu Lin"
Spike-HAR++: an energy-efficient and lightweight parallel spiking transformer for event-based human action recognition.
Front Comput Neurosci
November 2024
Article Synopsis
- Event-based cameras excel in human action recognition (HAR) due to their high dynamic range and efficiency, making them ideal for capturing fast movements.
- Spike Neural Networks (SNNs) are particularly effective with event camera data because they operate on a spike-driven paradigm, which offers lower power consumption than traditional neural networks.
- The paper introduces two innovative SNN models, Spike-HAR and Spike-HAR++, which enhance HAR accuracy through advanced spike attention mechanisms and efficient architecture, demonstrating impressive classification performance with minimal energy usage.
A bitter taste often identifies hazardous compounds and it is generally avoided by most animals and humans. Bitterness of hydrolyzed proteins is caused by the presence of bitter peptides. To improve palatability, bitter peptides need to be identified experimentally in a time-consuming and expensive process, before they can be removed or degraded.
View Article and Find Full Text PDF