Learning Globally-Consistent Local Distance Functions for Shape-Based Image Retrieval and Classification

By Andrea Frome, Yoram Singer, Fei Sha, Jitendra Malik. IEEE International Conference on Computer Vision, 2007

429 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/861839d67bacd1d2e5a956a172dddc8644e634fb

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

Large Scale Visual Recognition (2012, methodology): “Furthermore the proposed techniques are easily parallelizable, allowing better scaling than [105] (even without hierarchy) which already improved com3Closely related in technique to [108] for classification.”
Associative Embedding: End-to-End Learning for Joint Detection and Grouping (2016, methodology): “Works in image retrieval have used vector embeddings to measure similarity between images [12, 43].”
Action Recognition with Exemplar Based 2.5D Graph Matching (2012, background): “While the majority of work in computer vision are model based, exemplar based methods have also been applied in object recognition [25–27] and video classification [28].”
Segmentation-Aware Convolutional Networks Using Local Attention Masks (2017, background): “The goal of metric learning is to produce features from which one can estimate the similarity between pixels or regions in the input [18].”
Sparse distance learning for object recognition combining RGB and depth information (2011, background): “{kevinlai,lfb}@cs.washington.edu Xiaofeng Ren is with Intel Labs Seattle, Seattle, WA 98105, USA.”
Is that you? Metric learning approaches for face identification (2009, background): “It is not “local” in the sense of usual kNN classifiers or other “local learning” methods [4, 9], as MkNN measures the correspondence between two distinct local neighbourhoods.”
A Survey on Metric Learning for Feature Vectors and Structured Data (2013, background): “The work of Frome et al. (2007) is one of the first to propose to learn multiple local metrics.”
Gaussian Processes for Object Categorization (2010, methodology): “…various approaches based on support vector machines or nearest neighbor classiﬁcation (Grauman and Darrell 2005; Zhang et al. 2006; Wallraven et al. 2003; Nister and Stewenius 2006; Lazebnik et al. 2006; Varma and Ray 2007; Bosch et al. 2007; Frome et al. 2007; Kumar and Sminchisescu 2007).”
A Multimedia Retrieval Framework Based on Semi-Supervised Ranking and Relevance Feedback (2012, background): “In addition, researchers developed a specific distance metric for shape images [13].”
Fast Neighborhood Graph Search Using Cartesian Concatenation (2013, background): “It is also a crucial step in many vision and graphics problems, such as shape matching [13], object retrieval [29], feature matching [8, 35], texture synthesis [23], image completion [16] and so on.”
Hamming Distance Metric Learning (2012, methodology): “Such loss functions have been used in metric learning [5, 11], and, as shown below, they are also naturally suited to Hamming distance metric learning.”
SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmentation (2017, background): “Our work is also closely related to similarity metric learning, which has been widely used in deep learning on various tasks such as person re-identification [52], matching [16], image retrival [12, 50] and face recognition [6].”
Human Curation and Convnets: Powering Item-to-Item Recommendations on Pinterest (2015, methodology): “Visual features are widely used in both content-based recommendation systems and image search systems [5], [11], and the the learning-to-rank framework used in this paper from [14] has been widely used in industry [8], [9], [12], [18].”
Metric Learning: A Survey (2013, background): “(As considered in [23], however, each feature is a multi-dimensional vector obtained by a feature extractor on an image patch, and additionally a matching algorithm is applied to match features from pairs of images.”
Recognition using regions (2009, methodology): “With 15 and 30 training images per category, our method outperforms [ 14 ], [15], [33], [13] and [19] but not [5].”
In defense of Nearest-Neighbor based image classification (2008, neutral): “(b) Multiple descriptor type methods: ‘NBNN (5 Desc)’, ‘Bosch Trees’ (with ROI Optimization) [5], ‘Bosch SVM’ [6], ‘LearnDist’ [ 11 ], ‘SKM’ [15], ‘Varma’ [27], ‘KTA’ [18].”