Fast Human Attention Prediction for Fixation-guided Active Perception in Autonomous Navigation

2026-06-18

AI Summary

This work addresses the challenge of efficiently modeling human visual attention in robotic systems, where existing gaze prediction models are often computationally prohibitive for deployment in autonomous navigation. The authors propose GazeLNN, a lightweight autoregressive model for sequential gaze prediction that introduces Liquid Neural Networks to this task for the first time. Integrated with MobileNetV3 for visual feature extraction, GazeLNN recursively generates attention heatmaps conditioned on both the current image and prior fixations. Evaluated on the MIT low-resolution dataset, the model achieves a ScanMatch score of 0.47—significantly outperforming existing recurrent approaches—while reducing computational cost by 99.4% and accelerating inference by sixfold. Furthermore, GazeLNN is successfully incorporated into a reinforcement learning–based active perception strategy, demonstrating effective fixation-guided autonomous navigation on a real aerial robotic platform.

Abstract

Human visual attention relies on structured scanpaths to efficiently process scenes, yet instilling this behavior into robot autonomy is in its infancy and hindered by the high,computational costs of existing predictive models. To address this, we introduce GazeLNN, a computationally lightweight,scanpath prediction model that leverages Liquid Neural Networks as its recurrent engine and employs MobileNetV3 for feature extraction. Operating auto-regressively, the architecture predicts sequential fixation heatmaps conditioned on the current visual stimulus and fixation history. Despite requiring only 0.61 GFLOPs, GazeLNN achieves state-of-the-art performance on the MIT Low Resolution dataset achieving 0.47 ScanMatch score. It outperforms existing recurrent baselines across diverse evaluation metrics, while reducing computational costs by 99.40% and accelerating inference by up to six times. To investigate the role of human attention modeling in robot autonomy and demonstrate the practical utility of this highly efficient architecture, we integrate GazeLNN into an active camera-robot control policy trained via Reinforcement Learning. This integration enables human-fixation-guided perception during autonomous navigation, validated through successful real-world deployments on an aerial robot.

Problem

• human visual attention
• fixation prediction
• autonomous navigation
• active perception
• computational efficiency

Innovation

• GazeLNN
• Liquid Neural Networks
• scanpath prediction
• active perception
• autonomous navigation