Saliency-based Sequential Image Attention with Multiset Prediction
This addresses multi-label image classification with multiset prediction, enabling arbitrary label permutations and multiple instances per label, which is an incremental improvement over conventional models.
The paper tackles multi-label image classification by proposing a hierarchical visual architecture that uses a novel saliency-based sequential attention mechanism to focus on salient regions, achieving high precision and recall while localizing objects.
Humans process visual scenes selectively and sequentially using attention. Central to models of human visual attention is the saliency map. We propose a hierarchical visual architecture that operates on a saliency map and uses a novel attention mechanism to sequentially focus on salient regions and take additional glimpses within those regions. The architecture is motivated by human visual attention, and is used for multi-label image classification on a novel multiset task, demonstrating that it achieves high precision and recall while localizing objects with its attention. Unlike conventional multi-label image classification models, the model supports multiset prediction due to a reinforcement-learning based training process that allows for arbitrary label permutation and multiple instances per label.