Human motion analysis: a review
Computer Vision and Image Understanding
The Representation and Recognition of Human Movement Using Temporal Templates
CVPR '97 Proceedings of the 1997 Conference on Computer Vision and Pattern Recognition (CVPR '97)
Recognizing Human Actions: A Local SVM Approach
ICPR '04 Proceedings of the Pattern Recognition, 17th International Conference on (ICPR'04) Volume 3 - Volume 03
International Journal of Computer Vision
Vehicle Categorization: Parts for Speed and Accuracy
ICCCN '05 Proceedings of the 14th International Conference on Computer Communications and Networks
IEEE Transactions on Pattern Analysis and Machine Intelligence
LIBSVM: A library for support vector machines
ACM Transactions on Intelligent Systems and Technology (TIST)
Sampling strategies for bag-of-features image classification
ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part IV
Local descriptors for spatio-temporal recognition
SCVMA'04 Proceedings of the First international conference on Spatial Coherence for Visual Motion Analysis
Sparse coding on local spatial-temporal volumes for human action recognition
ACCV'10 Proceedings of the 10th Asian conference on Computer vision - Volume Part II
Hi-index | 0.00 |
We investigate the relative strengths of existing space-time interest points in the context of action detection and recognition. The interest point operators evaluated are an extension of the Harris corner detector (Laptev et al. [1]), a space-time Gabor filter (Dollar et al. [2]), and randomized sampling on the motion boundaries. In the first level of experiments we study the low level attributes of interest points such as stability, repeatability and sparsity with respect to the sources of variations such as actors, viewpoint and action category. In the second level we measure the discriminative power of interest points by extracting generic region descriptors around the interest points (1. histogram of optical flow[3], 2. motion history images[4], 3. histograms of oriented gradients[3]). Then we build a simple action recognition scheme by constructing a dictionary of codewords and learning a recognition system using the histograms of these codewords. We demonstrate that although there may be merits due to the structural information contained in the interest point detections, ultimately getting as many data samples as possible, even with random sampling, is the decisive factor in the interpretation of space-time data.