2022
Style transformed synthetic images for real world gaze estimation by using residual neural network with embedded personal identities
Wang Q, Wang H, Dang R, Zhu G, Pi H, Shic F, Hu B. Style transformed synthetic images for real world gaze estimation by using residual neural network with embedded personal identities. Applied Intelligence 2022, 53: 2026-2041. DOI: 10.1007/s10489-022-03481-9.Peer-Reviewed Original ResearchGaze directionNatural human-robot interactionPersonal identitySynthetic imagesHuman-robot interactionResidual neural networkNeural network structureSignificant performance gainsGaze estimationBenchmark datasetsTraining dataArt resultsNeural networkGaze interactionCurrent best resultsSocial communicationPerformance gainsReal worldNew personBetter resultsStyle transformationNetwork structureDatasetImagesNetwork
2021
Learning Oculomotor Behaviors from Scanpath
Li B, Nuechterlein N, Barney E, Foster C, Kim M, Mahony M, Atyabi A, Feng L, Wang Q, Ventola P, Shapiro L, Shic F. Learning Oculomotor Behaviors from Scanpath. 2021, 407-415. DOI: 10.1145/3462244.3479923.Peer-Reviewed Original ResearchEye-tracking applicationsOculomotor behaviorAutism spectrum disorderEye-tracking dataPre-trained modelsLarge model sizeSemi-supervised tasksContrastive learning taskSpectrum disorderLearning taskClassification taskBaseline approachesDownstream tasksPredictive codingFixation identificationRich representationTraining datasetModel sizeVariety of applicationsTaskScanpathsFramework modelAblation experimentsNovel methodBetter results