Author | Conference | Journal | Organization | Year | DOI
Look for results that meet for the following criteria:
Academic
Publications
Visual categorization with negative examples for free
Edit
Visual categorization with negative examples for free   (Citations: 4)
BibTex | RIS | RefWorks Download
Automatic visual categorization is critically dependent on labeled examples for supervised learning. As an alternative to traditional expert labeling, social-tagged multimedia is becoming a novel yet subjective and inaccurate source of learning examples. Different from existing work focusing on collecting positive examples, we study in this paper the po- tential of substituting social tagging for expert labeling for creating negative examples. We present an empirical study using 6.5 million Flickr photos as a source of social tag- ging. Our experiments on the PASCAL VOC challenge 2008 show that with a relative loss of only 4.3% in terms of mean average precision, expert-labeled negative examples can be completely replaced by social-tagged negative examples for consumer photo categorization.
Conference: ACM Multimedia Conference - MM , pp. 661-664, 2009
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
More »
    • ...Research on using data sources other than manually labelled samples for training concept classifiers has thus far focussed on the use of publicly available tagged multimedia resources, particularly images from Flickr [5, 6, 19, 26] and videos from YouTube [31–33]...
    • ...False negatives can also be filtered out by removing resources tagged with the concept’s name and its synonyms from the randomly selected negative samples [19]...
    • ...Concept classifiers trained with Flickr images have been applied both to test sets consisting of other Flickr images [6, 26 ]a nd across domain to test sets comprising images from the PASCAL VOC Challenge 5 [19] or TRECVID videos [5, 26]...
    • ... trained on tagged resources obtained from the Web is feasible, (ii) such classifiers achieve “fair” effectiveness when applied on data obtained from the same domain (i.e., other tagged resources) [6, 33], and (iii) such classifiers are outperformed in cross-domain settings by classifiers trained on manually annotated data obtained from the same domain as the target data, with the former though working well for some concepts [19, 31]...

    Theodora Tsikrika, et al. Reliability and effectiveness of clickthrough data for automatic image...

    • ...For example, the automatic generation of the training data is implemented by exploring external images from the web [4], Flickr images [6, 17, 27, 33], YouTube videos [35], and clickthrough data [34]...

    Zhineng Chen, et al. Web video retagging

    • ...Such efforts include the recent works in [8, 15]...
    • ...In [15], in view that as high as 90% of manual labeling efforts are spent on identifying negative samples, concept learning is conducted by direct collection of negative samples from user-tagged images, together with expert-labeled positive examples...
    • ...In addition, Li et al. [15] empirically shown that replacing expert-labeled negative examples with social tagged images for concept learning only results in slight loss of detection performance...

    Shiai Zhu, et al. On the sampling of web images for learning visual concept classifiers

    • ...Generally, Support Vector Machine (SVM) has been widely used as a semantic concept detector in the areas of image classification and retrieval [5-8]...
    • ...It is widely accepted that SVM may have different classification performances depending largely on how to organize (or construct) positive and negative training samples [8], [11]...
    • ...However, the negative samples of a particular concept detector are rather randomly, as opposed to the case for positive training samples, collected from the positive training samples used to form the rest of detectors [8]...
    • ...SVM is widely used to design the detectors in the area of image classification and retrieval using the detectors [4-6][8]...
    • ...However, the performance of SVM could be depending on how to organize (or construct) positive and negative training samples [8]...
    • ...In this subsection, we describe the conventional training strategy for semantic concept detectors which have been used in the area of image classification and retrieval [8]...
    • ...Testing data sets The conventional strategy [8] The proposed strategy...
    • ...The positive concepts The training strategy of detectors The conventional strategy [8] The proposed strategy...

    Jaehyun Jeon, et al. Training Strategy of Semantic Concept Detectors Using Support Vector M...

Order by:

Citations (4)

Share this on | Contribute to Academic