A Visual Category Filter for Google Images

By Rob Fergus, Pietro Perona, Andrew Zisserman. 2004

366 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Information Retrieval & Search

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 Information Retrieval & Search

Citations in the corpus

Webly Supervised Learning of Convolutional Networks (2015, background): “So, what we want is an approach that can learn visual representation from Flickr-like images.”
A sparse object category model for efficient learning and exhaustive recognition (2005, methodology): “Curves have been used to a limited extent in previous models for object categories, for example both Fergus et al. [7] and Jurie & Schmid [10] introduce curves as a feature type.”
Towards total scene understanding: Classification, annotation and segmentation in an automatic framework (2009, methodology): “Finally, our approach is inspired by earlier approaches of learning from noisy internet data such as [2, 19, 10, 11, 29].”
Animals on the Web (2006, neutral): “In addition, there has been some recent work on reranking Google search results using only images [11, 9] and on re-ranking search results using text plus images [21].”
A Boundary-Fragment-Model for Object Detection (2006, methodology): “[13] used boundary curves between bitangent points in their extension of the constellation model; and, Jurie and Schmid [16] detected circular arc features from boundary curves.”
One-shot learning of object categories (2006, methodology): “Alternative representations such as curve contours, which model the outline of the object, could also be used with little modification to the underlying model [14], [12].”
Exploiting weakly-labeled Web images to improve object classification: a domain adaptation approach (2010, methodology): “In order to overcome this limitation and in an attempt to buil d classifiers for arbitrary object classes, several authors have proposed systems that learn from weakl y-labeled Internet photos [10, 9, 29, 20].”
Sense discovery via co-clustering on images and text (2015, background): “There have been early efforts on automatically training visual classifiers using web data to build large datasets automatically [26, 39, 42, 27, 20], find iconic images [5, 35] and improve image retrieval results [19, 28, 45, 34].”
Towards Scalable Dataset Construction: An Active Learning Approach (2008, methodology): “[7, 8] re-ranked images obtained from Google Image Search using visual information.”
NEIL: Extracting Visual Knowledge from Web Data (2013, methodology): “A common way of automatically creating a dataset is to use image search results and rerank them via visual classifiers [14] or some form of joint-clustering in text and visual space [2] [34].”
Video search reranking via information bottleneck principle (2006, methodology): “Besides, authors of [5] used the Google image search return set as (pseudo-)positives and utilized a parts-based approach to learn the object model and then used that to rerank the search baseline images.”
Contour-based learning for object detection (2005, background): “(a) 0 0.1 0.2 0.3 0.4 0.5 0.5 0.6 0.7 0.8 0.9 1 1 − Precision Recall Full Model (K 2 ) K 2 (50 segmented) K 2 (no refinement) K 2 (no orientation) K 2 (no clutter training) Stage 1 only (K 1 ) K 2 (no perturbation) (b) 0 0.2 0.4 0.6 0.8 1 0.5 0.6 0.7 0.8 0.9 1 1−Precision Recall Our algorithm Fergus”
Harvesting large-scale weakly-tagged image databases from the web (2010, background): “Some pioneering works have been done to leverage Internet images for computer vision tasks [2, 4-8].”
Content-based visual search learned from social media (2012, methodology): “Quite a few methods follow this research line, either by reranking search results in light of visual consistency [33,45,46,52,90,139] or by expanding the original queries [10, 21, 72, 86].”
80 Million Tiny Images: A Large Data Set for Nonparametric Object and Scene Recognition (2008, background): “However, web images define an interesting visu al world for developing computer vision applications [16], [3 7].”
SalientShape: group saliency in image collections (2013, background): “We found that considering the largest appearance cluster [3] or top-ranked internet images [25] as an initial set to be unsatisfactory.”
How flickr helps us make sense of the world: context and content in community-contributed media collections (2007, background): “The SIFT features have been found to be extremely powerful in a number of vision applications, from generic object recognition [7] to alignment and registration between various photographs of a single scene [22].”
Finding iconic images (2009, methodology): “The goals of both these papers and our own differ significantly from that of a body of related work on re-ranking the results of a search engine (usually Google) [10, 9, 3, 19, 7].”
Dataset Issues in Object Recognition (2006, background): “Berg and Forsyth consider in [7] the problem of combining the text and image information stored on Web pages to re-rank Google search results for a set of queries (see [11,32] for related work).”
50 Years of object recognition: Directions forward (2013, background): “Fergus, Perona and Zisserman are arguably some of the strongest advocates of the approach and have published a series of papers on constellation methods, some of which we overview here [100], [195], [174], [200], [203].”