Learning from ambiguously labeled images

By Timothée Cour, Benjamin Sapp, Christopher T. Jordan, B. Taskar. 2009

204 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

Citations in the corpus

End-to-End Learning of Visual Representations From Uncurated Instructional Videos (2019, methodology): “The method in [8] presents an alternative solution by formulating the association problem using a convex surrogate loss.”
A dataset for Movie Description (2015, background): “Their corpus is based on the original sources to create the DVS and contains different kinds of artifacts not present in actual description, such as dialogs and production notes.”
Speech2Action: Cross-Modal Supervision for Action Recognition (2020, background): “Aligning Screenplays to Movies: A number of works have explored the use of screenplays to learn and automatically annotate character identity in TV series [5, 9, 30, 35, 39].”
Learning from Partial Labels (2011, background): “A preliminary version of this work appeared in Cour et al. (2009).”
Learning from Video and Text via Large-Scale Discriminative Clustering (2017, methodology): “The method in [8] presents an alternative solution by formulating the association problem using a convex surrogate loss.”
Label Ranking under Ambiguous Supervision for Learning Semantic Correspondences (2010, result): “Cour et al. (2009) propose a proxy risk for multiclass classification under ambiguous supervision, and prove a result similar in essence to Theorem 3 for that case.”
Detecting People Looking at Each Other in Videos (2013, background): “…is another step towards a fuller video understanding, and complements recent work on automatic character identi-ﬁcation (Everingham et al. 2006; Cour et al. 2009; Sivic et al. 2009), human pose estimation (Ferrari et al. 2009; Andriluka et al. 2009; Bourdev et al. 2010; Sapp et al. 2010;Yang…”
Condensed Movies: Story Based Retrieval with Contextual Embeddings (2020, background): “Early works, however, focused on using film and TV to learn human identity [21,22,23,24,25,26] or human actions [27,28,29,30,31] from the scripts or captions accompanying movies.”
Finding Actors and Actions in Movies (2013, methodology): “Previous work has explored video scripts to learn and automatically annotate characters in TV series [6, 22, 23].”
Linking People in Videos with "Their" Names Using Coreference Resolution (2014, methodology): “First, note that even our unidirectional model (OurUnidir) performs better than the state-of-the-art methods from [6, 7].”
Aligning plot synopses to videos for story-based retrieval (2015, background): “The recent advances in this domain are mainly in popular computer vision topics such as action recognition [21], person identification [6,10,37], pose estimation [5], and human interaction analysis [30].”
It ’ s in the bag : Stronger supervision for automated face labelling (2015, methodology): “This topic has received plenty of attention in the community and papers have investigated the problem under various assumptions on the available information [2, 3, 5, 4, 6, 8, 9, 12, 17, 18, 20, 22, 24].”
A data-driven approach to cleaning large face datasets (2014, methodology): “This means we cannot exploit useful information such as co-occurrences to help disambiguate their identities, as was done in [6, 8, 9].”
"Here's looking at you, kid". Detecting people looking at each other in videos (2011, background): “Determining whether characters are interacting using their eyelines is another step towards a fuller video understanding, and complements recent work on automatic character identification [7, 11, 27], human pose estimation [1, 6, 9, 13, 14, 24], human action recognition [16, 17, 18], and specific in”
Book2Movie: Aligning video scenes with book chapters (2015, background): “Fan transcripts with subtitles are emerging as the de facto standard in automatic person identification [6, 8, 19, 24] and action recognition [5, 12].”