Efficient Object Category Recognition Using Classemes

By Lorenzo Torresani, Martin Szummer, Andrew William Fitzgibbon. 2010

464 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Learning Visual Features from Large Weakly Supervised Data (2015, methodology): “Tor-resani et al. [49] train a collection of weak object classiﬁers and use the classiﬁer outputs as additional image features (in addition to low-level image features).”
Webly Supervised Learning of Convolutional Networks (2015, background): “So, what we want is an approach that can learn visual representation from Flickr-like images.”
DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition (2013, methodology): “…approach proven successful in computer vision and multimedia settings using a concept-bank paradigm (Kennedy & Hauptmann, 2006; Li et al., 2010; Torresani et al., 2010) by learning the features on large-scale data in a supervised setting, then transferring them to different tasks with…”
Unsupervised Discovery of Mid-Level Discriminative Patches (2012, neutral): “[15,3,16]).”
Write a Classifier: Predicting Visual Classifiers from Unstructured Text (2015, methodology): “) is extracted, a subset of feature dimensions was selected [55], and a one-versus-all classifier ’c is trained for each category.”
Video Event Detection by Inferring Temporal Instance Labels (2014, background): “One focus of previous research is on designing new features, including low-level features of visual features [6, 12], action features [22], audio features [14], and mid-level representation including concept feature, attributes [20] etc.”
Model recommendation: Generating object detectors from few samples (2015, background): “In previous work, examples of transfer of prior experience to a new task include concept drift [51], domain adaption, transfer learning [40] (e.g., sparse prototype representations [43], hypothesis transfer learning [27, 49], regularized SVM [4]), multi-task learning (e.g., rank-reduced shared class”
Iterative Quantization: A Procrustean Approach to Learning Binary Codes for Large-Scale Image Retrieval (2013, unknown): “As an additional application, Section 7 demonstrates the usefulness of ITQ for learning binary visual attributes or " classemes " [46] on the ImageNet dataset [8].”
Enhanced representation and multi-task learning for image annotation (2013, background): “One promising approach to transfer knowledge from some classes to others is to share information between object classes on the level of individual features [10, 11, 12, 13, 14] or entire object class models [15, 16].”
How should we evaluate supervised hashing? (2016, background): “This is consistent with prior observations [17] that attribute vectors from arbitrary classes are an efficient representation for global image matching, and further extends the Fig.”
Harnessing Lab Knowledge for Real-World Action Recognition (2014, result): “This idea is similar to the classemes feature proposed in Torresani et al. (2010).”
A codebook-free and annotation-free approach for fine-grained image categorization (2012, methodology): “While these approaches have shown promising results in basic-level object and scene classification tasks, they still suffer from the issues of coarse codebook encoding ([6] where the best results are obtained by clustering image regions) or tedious human annotation ([18, 25, 19]).”
Semantic Aware Video Transcription Using Random Forest Classifiers (2014, background): “[29] showed that object detector responses provide competitive performance when used as features for image classification task.”
Learning Attributes Equals Multi-Source Domain Generalization (2016, background): “Attributes also reveal the rich structures underlying categories and are thus often employed to regulate machine learning models for visual recognition [69, 73, 45, 63, 33, 22, 21, 62].”
Exploiting weakly-labeled Web images to improve object classification: a domain adaptation approach (2010, background): “This binarization has been shown to yield very little degradation in recognition p erformance (see [28] for further details).”
High-dimensional signature compression for large-scale image classification (2011, background): “Other works have proposed to perform a semantic dimensionality reduction [31, 28].”
Understanding Objects in Detail with Fine-Grained Attributes (2014, background): “Thus, learning attributes from global object or image cues as commonly done [8, 9, 13, 14, 21, 25, 30, 33– 36, 39, 42–44] may in fact pick up contextual evidence more than learning about the attribute per se.”
Zero-Shot Visual Recognition Using Semantics-Preserving Adversarial Embedding Networks (2017, unknown): “, visual attributes “striped tail” [13, 27, 67], classeme “fox-like” [57, 30, 65, 52], or “raccoon” word vectors [47, 37].”