Glance

Research behind the predictions

Published research supplies the methods. Each asset keeps its own license.

Credits below describe planned use, not proof that a model has shipped. Code licenses do not automatically cover weights, datasets or paper text. Complete bibliography and BibTeX: CITATIONS.md at the repository root.

Paper, model and dataset references
Title / assetAuthorsInstitutionVenue / yearLinkLicenseGlance use
A Model of Saliency-Based Visual Attention for Rapid Scene AnalysisLaurent Itti; Christof Koch; Ernst NieburCaltech; additional affiliation TODO-verifyPAMI 1998Paper / DOIPaper/implementation terms TODO-verifyConceptual saliency and inhibition-of-return background; no original assets bundled.
Learning to Predict Where Humans Look / MIT1003Tilke Judd; Krista Ehinger; Frédo Durand; Antonio TorralbaMIT; additional affiliation TODO-verifyICCV 2009Paper / DOIPaper and dataset terms TODO-verify; not mirroredHistorical saliency dataset and prediction reference; no dataset distribution.
What do different evaluation metrics tell us about saliency models?Zoya Bylinskii; Tilke Judd; Aude Oliva; Antonio Torralba; Frédo DurandMITPAMI 2019; arXiv 2016Paper / DOIPaper reuse terms TODO-verifyMetric interpretation; AUC, CC and NSS explanations.
DeepGaze IIE: Calibrated prediction in and out-of-domain for state-of-the-art saliency modelingAkis Linardos; Matthias Kümmerer; Ori Press; Matthias BethgeUniversity of TübingenICCV 2021Paper / DOINo model redistribution license established in SPEC; user-import only; paper terms TODO-verifyReference benchmark values only; imported ONNX support subject to verification.
Predicting Visual Importance Across Graphic Design Types / UMSI / Imp1kCamilo Fosco; Vincent Casser; Amish Kumar Bedi; Peter O’Donovan; Aaron Hertzmann; Zoya BylinskiiMIT + Adobe; full affiliation mapping TODO-verifyUIST 2020Paper / DOIPaper, code, weights and dataset terms TODO-verifyDesign-domain importance background; not bundled or claimed as the SUM model.
UEyes: An Eye-Tracking Dataset across User Interface TypesYue Jiang; Luis A. Leiva; Paul R. B. Houssel; Hamed R. Tavakoli; Julia Kylmälä; Antti OulasvirtaAalto University; full affiliations TODO-verifyCHI 2023 workshop; arXiv 2024Paper / DOIPaper and dataset terms TODO-verify; not mirroredUI-domain gaze dataset reference; workshop derivative distinguished from main research paper.
EyeFormer: Predicting Personalized Scanpaths with Transformer-Guided Reinforcement LearningYue Jiang; Zixin Guo; Hamed R. Tavakoli; Luis A. Leiva; Antti OulasvirtaAalto University; University of Luxembourg; Nokia TechnologiesUIST 2024Paper / DOINo model license established in SPEC; user-import only; paper terms TODO-verifyOptional UI scanpath import; excluded from redistributed assets.
SUM: Saliency Unification through Mamba for Visual Attention ModelingAlireza Hosseini; Amirhossein Kazerouni; Saeed Akhavan; Michael Brudno; Babak TaatiUniversity of Tehran; University of Toronto; Vector Institute; University Health NetworkWACV 2025Paper / DOIMIT code per SPEC; weights/dependencies and paper terms TODO-verifyCandidate image model. ONNX selective-scan export and parity pending; exclude from v1 if unsupported.
Unified Image and Video Saliency Modeling / UNISALRichard Droste; Jianbo Jiao; J. Alison NobleUniversity of OxfordECCV 2020Paper / DOIApache-2.0 code per SPEC; weights/dependencies and paper terms TODO-verifyCandidate image/video model; ONNX export and parity pending.
TranSalNet: Towards perceptually relevant visual saliency predictionLou et al.; full author list TODO-verifyCardiff University; full affiliations TODO-verifyNeurocomputing 2022Paper / DOIMIT code per SPEC; weights/dependencies and paper terms TODO-verifyCandidate image model; ONNX export and parity pending.
CAT2000: A Large Scale Fixation Dataset for Boosting Saliency ResearchAli Borji; Laurent IttiUniversity of Southern California; affiliation mapping TODO-verifyarXiv 2015Paper / DOIPaper and dataset terms TODO-verify; not mirroredBenchmark context; dataset linked, never redistributed.
MIT/Tübingen Saliency Benchmark / MIT300Benchmark contributors; full author attribution TODO-verifyMIT + University of TübingenBenchmark website; year TODO-verifyPaper / DOIWebsite/data terms TODO-verify; not mirroredSource for supplied DeepGaze IIE and human reference values. Exact leaderboard row/protocol pending.
Saliency Benchmarking Made Easy: Separating Models, Maps and MetricsMatthias Kümmerer; Thomas S. A. Wallis; Matthias BethgeUniversity of Tübingen; full affiliations TODO-verifyECCV 2018Paper / DOIPaper terms TODO-verifyBenchmark protocol background; model density and metric-specific maps differ.

Redistribution boundary

SUM and TranSalNet have MIT code licenses in SPEC; UNISAL has Apache-2.0. Weight and dependency rights require verification. EyeFormer and DeepGaze IIE remain user-import only under SPEC’s no-license policy. Dataset files are not mirrored.

SUM ONNX export and selective-scan support are unverified. If export cannot pass parity checks, exclude it from v1 and record the outcome in CITATIONS.md.

Questions before you start

Does Glance record my eyes?

No. Glance predicts likely attention from an image. It does not record gaze or use a camera.

Is Glance free?

Glance is a free, open-source desktop app in development. Its own code is MIT; model assets have separate terms.