Online domain adaptation of a pre-trained cascade of classifiers

Jain, Virendra; Learned-Miller, Erik

doi:10.1109/cvpr.2011.5995317

Cited by 154 publications

(99 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Unlike other domains such as speech recognition and handwriting recognition, where adaptation has been indispensable, adaptation for visual object detection has received relatively little attention. Some early work has been conducted in this area [219,220,202,203] and we strongly believe that this is a great direction for future work. …”

Section: Discussion Future Work and Conclusionmentioning

confidence: 99%

“…For example, how do current face detectors trained on standard databases perform in digitized images of the 19th century or in a collection of out-of-focus images from surveillance cameras for a particular environment? Instead of retraining the detectors from scratch when a new collection is available, an interesting topic of further research is to develop techniques which can adapt to a new image dataset without having access to the original training data [202]. That way camera and environment specific face detectors with very high performance could be routinely developed.…”

Section: Discussion Future Work and Conclusionmentioning

confidence: 99%

See 1 more Smart Citation

A survey on face detection in the wild: Past, present and future

Zafeiriou

Cha

Zhang

2015

Computer Vision and Image Understanding

386

167

View full text Add to dashboard Cite

Section: Discussion Future Work and Conclusionmentioning

confidence: 99%

Section: Discussion Future Work and Conclusionmentioning

confidence: 99%

A survey on face detection in the wild: Past, present and future

Zafeiriou

Cha

Zhang

2015

Computer Vision and Image Understanding

386

167

View full text Add to dashboard Cite

“…Hence, our main emphasis is to prove their advantage over Haar-like features that is clearly shown in Figure 9(b). Computational more complex methods like the results of Li et al [24] and Jain et al (VJGPR) [25] shown in Figure 9(b) use SURF features [24] or model inter-detection dependencies to exploit scene information [25] to derive higher detection rates. But especially Li et al [24] propose besides SURF features a AUC score for cascade training as a second contribution that also adds to their good results and is not in contrast to 2Rec-Features.…”

Section: Face Detectionmentioning

confidence: 99%

Exploiting Object Characteristics Using Custom Features for Boosting-Based Classification

2013

View full text Add to dashboard Cite

Abstract. Typical feature pools used to train boosted object detectors contain various redundant and unspecific information which often yield less discriminative detectors. In this paper we introduce a feature mining algorithm taking domain specific knowledge into account. Our proposed feature pool contains rectangular shaped features generated from an image clustering algorithm applied on the mean image of the object training set. A combination of two such spatially separated rectangular regions yields a set of features which have a similar evaluation time like classical Haar-like features, but are much smarter (automatically) selected and more discriminative since image correlations can be more consequently exploited. Overall, training is faster and results in more selective detectors showing improved precision. Several experiments demonstrate the gain when using our proposed feature set in contrast to standard features.

show abstract

“…(Note: Background regions have been obfuscated for legal/privacy reasons) problem of domain adaptation, seeking to develop effective mechanisms to transfer or adapt knowledge from one domain to another related domain. While these advances have also been applied by the computer vision community with promising results [22,17,15,1], object models are still being trained and tested on images consisting of only one object zoomed and cropped at the center of a relatively uniform background. As a result, in such experimental settings the general problem of object detection is reduced to that of image classification.…”

Section: Introductionmentioning

confidence: 99%

Domain adaptive object detection

Mirrashed

Morariu

Siddiquie

et al. 2013

2013 IEEE Workshop on Applications of Computer Vision (WACV)

View full text Add to dashboard Cite

We study the use of domain adaptation and transfer learning techniques as part of a framework for adaptive object detection. Unlike recent applications of domain adaptation work in computer vision, which generally focus on image classification, we explore the problem of extreme class imbalance present when performing domain adaptation for object detection. The main difficulty caused by this imbalance is that test images contain millions or billions of negative image subwindows but just a few image subwindows containing positive instances, which makes it difficult to adapt to changes in the positive classes present new domains by simple techniques such as random sampling. We propose an initial approach to addressing this problem and apply our technique to vehicle detection in a challenging urban surveillance dataset, demonstrating the performance of our approach with various amounts of supervision, including the fully unsupervised case.

show abstract

Online domain adaptation of a pre-trained cascade of classifiers

Abstract: Abstract

Cited by 154 publications

References 11 publications

A survey on face detection in the wild: Past, present and future

A survey on face detection in the wild: Past, present and future

Exploiting Object Characteristics Using Custom Features for Boosting-Based Classification

Domain adaptive object detection

Contact Info

Product

Resources

About