Learning object detection from a small number of examples: the importance of good features

By Kobi Levi, Yair Weiss. Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004., 2004

459 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/23dc97e672467abe20a4e9d5c6367a18f887dcd1

How researchers see this paper

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

Citations in the corpus

On-line Boosting and Vision (2006, methodology): “In this paper we use the standard Haar-like features from Viola and Jones [31] and in addition orientation histograms (with 16 bins) similar to [13, 6] and a simple version (4-th neighborhood) of local binary patterns (LBP) [17] as features.”
A survey on face detection in the wild: Past, present and future (2015, background): “In [80] local edge orientation histograms was proposed, which compute the histogram of edge orientations in subregions of the test windows.”
Ensemble Tracking (2005, background): “The histogram of oriented gradients is easy to compute and conveys rich information that was used in the past for detection and recognition purposes [11], [19], [10].”
Human detection using depth information by Kinect (2011, methodology): “Some methods involve statistical training based on local features, e.g. gradient-based features such as HOG [1], EOH [8], and some involve extracting interest points in the image, such as scale-invariant feature transform (SIFT) [9], etc.”
Cross-generalization: learning novel classes from a single example by feature replacement (2005, background): “Typically, hundreds of examples are required for training by the current classification algorithms [1] [9] [13] [15] [17] [19].”
Recursive Feature Generation for Knowledge-based Learning (2018, background): “In many real-world applications, however, the given set of features is insufﬁcient for inducing a high quality classiﬁer (Levi & Weiss, 2004; Paulheim & F¨umkranz, 2012).”
Multi-target tracking by on-line learned discriminative appearance models (2010, unknown): “, Haarlike features [24], SIFT-like features [12, 19], orientation histograms [10, 17], spatiograms [6], are integrated in this framework.”
Sketch-based manga retrieval using manga109 dataset (2015, methodology): “Note that integral images are utilized to enable the fast computing of features in the same manner as in [2].”
Iterative Hierarchical Optimization for Misspecified Problems (IHOMP) (2016, background): “6 Why are problems misspeciﬁed? While a problem description can be misspeciﬁed for many reasons, one important case is due to the state representation. It is well established in the machine learning [Levi and Weiss, 2004, Zhou et al., 2009] and RL [Konidaris et al., 2011] literature that “good” feat”
DSFD: Dual Shot Face Detector (2018, background): “Feature Learning Early works on face detection mainly rely on hand-crafted features, such as Harr-like features [29], control point set [1], edge orientation histograms [13].”
Cat Head Detection - How to Effectively Exploit Shape and Texture Features (2008, methodology): “Levi and Weiss [7] proposed three kinds of features on the oriented gradients: EOH1(R, k1, k2) = (S1(R) + )/(S2(R) + ), EOH2(R, k) = (S(R) + )/( ∑ j(S (R) + )), EOH3(R, R, k) = (S(R) − S(R))/sizeof(R),”
Human detection from images and videos: A survey (2016, methodology): “For example, edge orientation histogram (EOH) proposed by Levi and Weiss [89] together with rectangular features [158] were used in [17].”
Survey of Pedestrian Detection for Advanced Driver Assistance Systems (2010, methodology): “Recently, Geró nimo et al. [30] made use of Real AdaBoost to select the best features among a set of Haar-like and edge orientation histograms (EOHs; [89]) features to classify ROIs in the VS. EOHs first compute the gradient magnitude of the image, and then, distribute the pixels into k different…”
A Survey of Recent Advances in Face Detection (2010, methodology): “Statistics-based features Edge orientation histograms [45, 13] etc.”
Person Re-identification by Descriptive and Discriminative Classification (2011, background): “, Haar-like [21], Edge Orientation Histograms [12], or boundary fragments [15] have been introduced for the application with boosting for feature selection.”
Real-Time Human Detection Using Relational Depth Similarity Features (2010, methodology): “There has been much research in the past into human detection, and various different methods have been proposed[1][2][3][4][5].”
Learning to Learn: Model Regression Networks for Easy Small Sample Learning (2016, background): “In particular, inter-class transfer [40] and crossgeneralization [7] are achieved by discovering shared feature representations: 1) captured by linear or nonlinear feature transformations [14, 63, 85, 31, 1, 94, 48], 2) obtained by feature selection [59, 60, 27] or regularization [37], 3) described ”
Toward Practical Smile Detection (2009, methodology): “c. Edge Orientation Histograms (EOH): These features have recently become popular for a wide variety of tasks, including object recognition (e.g., in SIFT [32]) and face detection [ 33 ].”