Brain-Inspired Attention Model for Object Counting

Sequential visual attention and reinforcement learning for accurate object counting.

Authors: Abhijeet Sinha, Sweta Kumari, and V. Srinivasa Chakravarthy
Venue: International Conference on Neural Information Processing (ICONIP), 2023

This work presents a brain-inspired sequential attention model that uses Q-learning and recurrent visual glimpses for object counting. By learning where to look and integrating information across glimpses, the model achieved 92.1% object-counting accuracy.

Read the published paper.