Exploiting Neighbors for Faster Scanning Window Detection in Images - Advanced Concepts for Intelligent Vision Systems - page 224

For the neighborhood suppression to be useful, it must provide higher speed
than simple detector for the same precision of detection. To validate this, we have
trained number of detectors with different speeds (in terms of average number
of features computed per position) for each feature type. Then, we learned three
suppression classifiers with α set to 0.01, 0.05 and 0.2 for each of the detec-
tors. The corresponding speeds and detection rates are shown in Figure 5. Even
thought, only a single suppression classifier is used in this case for each of the de-
tectors, the results clearly show that by using neighborhood suppression, higher
speed can be reached for the same detection rate.
4 Conclusions
This paper presents a novel approach to acceleration of object detection through
scanning windows by prediction of the neighbor positions results using new clas-
sifiers that reuse the image features of the detector. The approach has been
demonstrated on frontal face and eye detection using WaldBoost classifiers. The
results clearly show that the proposed approach is feasible and that it can sig-
nificantly speed up the detection process without loss of detection performance.
Further work includes evaluation of the approach on further data sets, other
features, and possibly also different classification mechanisms, such as SVM.
Further work will also focus on real-time implementation of the proposed method
on CPU, GPU, and programmable hardware (FPGA). Also of interest will be
possible improved image scanning patterns that can benefit even more from the
neighborhood suppression.
Acknowledgements
This work has been supported by the Ministry of Education, Youth and Sports of
the Czech Republic under the research program LC-06008 (Center for Computer
Graphics) and by the research project ”Security-Oriented Research in Informa-
tional Technology” CEZMSMT, MSM0021630528.
References
1. Bourdev, L., Brandt, J.: Robust object detection via soft cascade. In: CVPR (2005)
2. Cha, Z., Viola, P.: Multiple-instance pruning for learning ecient cascade detectors.
In: NIPS (2007)
3. Dalal, N., Triggs, B.: Histograms of oriented gradients for human detection. In:
Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision
and Pattern Recognition (CVPR 2005), Washington, DC, USA, vol. 1, pp. 886-893.
IEEE Computer Society Press, Los Alamitos (2005)
4. Hou, C., Ai, H.Z., Lao, S.H.: Multiview pedestrian detection based on vector boost-
ing. In: Yagi, Y., Kang, S.B., Kweon, I.S., Zha, H. (eds.) ACCV 2007, Part I. LNCS,
vol. 4843, pp. 210-219. Springer, Heidelberg (2007)