Dynamic Scene Classification with Slow Features Analysis
Christian Theriault   ,   Nicolas Thome   ,   Matthieu Cord
Universite Pierre et Marie Curie, UPMC-Sorbonne Universities, LIP6, Paris, France
1.HMAX-S: Deep scale representation for biologically inspired image categorization. [pdf]:
Christian Theriault, Nicolas Thome, Matthieu Cord. ICIP 2011, p 1261-1264, ISBN: 978-1-4577-1304-0, Brussels, 11-14 Sep 2011
2.Extended coding and pooling in the HMAX model.
Christian Theriault, Nicolas Thome, Matthieu Cord. IEEE TRANSACTIONS ON IMAGE PROCESSING, Feb 2013, 22(2),764-777
3. Dynamic Scene Classification: Learning Motion Descriptors with Slow Features Analysis.
Christian Theriault, Nicolas Thome, Matthieu Cord. IEEE CVPR,2013, (with erratum below)
4. Perceptual principles for video classification with Slow Feature Analysis.
Christian Theriault, Nicolas Thome, Matthieu Cord, Patrick Pérez. IEEE Journal of Selected Topics in Signal Processing, 2014
[click here to download MATLAB code for ref 1-2]
[click here to download technical report including an erratum for ref 3]
Code for ref 3-4 available on demand or questions at:
This model is concerned with the learning of motion features for dynamic scenes classification based of slow feature analysis. The model takes inputs from basic V1 features as described here IP2013. From the V1 outputs of video frames the model learns motion features based of slow feature analysis (SFA) as illustrated below and thoroughly explained in this [technical report] .
Classification on two dynamic scenes data sets generates state of the art results.