Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Saliency Prediction in Monitoring Tasks: The Role of Visual Highlights

Zekun Wu, Anna Maria Feit

TL;DR

This paper investigates how visual highlights influence gaze during drone monitoring tasks and introduces the highlight informed saliency model (HISM) that fuses spatial context via ResNet50 with temporal cues from an AOI highlight sequence processed by an LSTM. HISM predicts saliency over time for highlighted AOIs and outperforms state of the art models SimpleNet and TASED-Net with a lower MSE. The experiments show that highlights accelerate initial attention to critical areas, supporting their use to improve situation awareness and prediction of gaze behavior. The work suggests practical impact for designing attention guiding interfaces and informs future saliency models to incorporate highlight cues.

Abstract

This study examines the role of visual highlights in guiding user attention in drone monitoring tasks, employing a simulated interface for observation. The experiment results show that such highlights can significantly expedite the visual attention on the corresponding area. Based on this observation, we leverage both the temporal and spatial information in the highlight to develop a new saliency model: the highlight-informed saliency model (HISM), to infer the visual attention change in the highlight condition. Our findings show the effectiveness of visual highlights in enhancing user attention and demonstrate the potential of incorporating these cues into saliency prediction models.

Enhancing Saliency Prediction in Monitoring Tasks: The Role of Visual Highlights

TL;DR

This paper investigates how visual highlights influence gaze during drone monitoring tasks and introduces the highlight informed saliency model (HISM) that fuses spatial context via ResNet50 with temporal cues from an AOI highlight sequence processed by an LSTM. HISM predicts saliency over time for highlighted AOIs and outperforms state of the art models SimpleNet and TASED-Net with a lower MSE. The experiments show that highlights accelerate initial attention to critical areas, supporting their use to improve situation awareness and prediction of gaze behavior. The work suggests practical impact for designing attention guiding interfaces and informs future saliency models to incorporate highlight cues.

Abstract

This study examines the role of visual highlights in guiding user attention in drone monitoring tasks, employing a simulated interface for observation. The experiment results show that such highlights can significantly expedite the visual attention on the corresponding area. Based on this observation, we leverage both the temporal and spatial information in the highlight to develop a new saliency model: the highlight-informed saliency model (HISM), to infer the visual attention change in the highlight condition. Our findings show the effectiveness of visual highlights in enhancing user attention and demonstrate the potential of incorporating these cues into saliency prediction models.
Paper Structure (4 sections, 3 figures)

This paper contains 4 sections, 3 figures.

Table of Contents

  1. Motivation and Study Design
  2. Visual Highlights' Influence on Gaze Behavior
  3. Highlight Informed Saliency Prediction
  4. Conclusions and Future Work

Figures (3)

  • Figure 1: Hit Rates and Response Time in Different Highlight Conditions
  • Figure 2: Highlight Informed Saliency Model Framework, which includes three key components: the ResNet50 to extract the spatial information, the two-layer LSTM block to generate the temporal features, and the MLP takes both types of features to predict the saliency of the AO
  • Figure 3: Normalized saliency of the critical AOI over time computed over 0.5s windows. The corresponding icon was highlighted at second 5. Our HISM model captures the time to first fixation of about 1s and the following decrease in attention. State-of-the-art image- and video-based saliency models (SimpleNet and TASED-Net) could not capture the temporal changes in users' attention