Simulating Human Gaze with Neural Visual Attention
This work addresses the need for more accurate human attention models in computer vision, though it appears incremental by building on existing attention modeling approaches.
The paper tackled the problem of modeling human visual attention with task guidance by proposing the Neural Visual Attention (NeVA) algorithm, which outperforms state-of-the-art unsupervised models in generating human-like scanpaths on three benchmark datasets.
Existing models of human visual attention are generally unable to incorporate direct task guidance and therefore cannot model an intent or goal when exploring a scene. To integrate guidance of any downstream visual task into attention modeling, we propose the Neural Visual Attention (NeVA) algorithm. To this end, we impose to neural networks the biological constraint of foveated vision and train an attention mechanism to generate visual explorations that maximize the performance with respect to the downstream task. We observe that biologically constrained neural networks generate human-like scanpaths without being trained for this objective. Extensive experiments on three common benchmark datasets show that our method outperforms state-of-the-art unsupervised human attention models in generating human-like scanpaths.