Project details for ABACOC Adaptive Ball Cover for Classification

Logo ABACOC Adaptive Ball Cover for Classification 2.0

by kikot - May 29, 2015, 11:57:28 CET [ BibTeX BibTeX for corresponding Paper Download ]

view (13 today), download ( 5 today ), 3 subscriptions



We introduce an online action recognition system that can be combined with any set of frame-by-frame (or local) feature descriptors. Our system covers the frame feature space with classifiers whose distribution adapts to the hardness of locally approximating the Bayes optimal classifier. An efficient nearest neighbour search is used to find and combine the local classifiers that are closest to the frames of a new video to be classified. The advantages of our approach are: incremental training, frame by frame real-time prediction, nonparametric predictive modelling, video segmentation for continuous action recognition, no need to trim videos to equal lengths.

Please Cite:

@inproceedings{de2014online, title={Online action recognition via nonparametric incremental learning}, author={De Rosa, Rocco and Cesa-Bianchi, Nicol{`o} and Gori, Ilaria and Cuzzolin, Fabio}, booktitle={Submitted to British Machine Vision Conference (BMVC 2014)}, year={2014} }

Changes to previous version:

version 2: parameterless system, constant model size, prediction confidence (for active learning)

BibTeX Entry: Download
Corresponding Paper BibTeX Entry: Download
Supported Operating Systems: Linux, Windows
Data Formats: Matlab, Java
Tags: Active Learning, Online Learning, Action Recognition, Incremental Learning, Multivariate Time Series Classification, Temporal Classification, Human Activity Recognition, Constant Model Size, Real Tim
Archive: download here

Other available revisons

Version Changelog Date

version 2: parameterless system, constant model size, prediction confidence (for active learning)

May 29, 2015, 11:57:28

Initial release of the library, future changes will be advertised shortly.

July 14, 2014, 16:27:03


No one has posted any comments yet. Perhaps you'd like to be the first?

Leave a comment

You must be logged in to post comments.