Журнал «Труды Института системного анализа Российской академии наук» - D.P. Matalov Quantile based soft cascade calibration

This paper presents an algorithm based on the Multiple Instance Pruning for calibrating the Viola and Jones soft cascades. The effectiveness of the proposed algorithm is showed in the practical problem of training detector of the 3rd page of a passport of a citizen of the Russian Federation in digital images obtained in uncontrolled conditions. The resultant detector showed a significantly lower false negative rate and the same false positive rates compared to the original Multiple Instance Pruning algorithm. To achieve that, the algorithm

follows statistical approach by considering the distributions of classifier responses on the entire set of acceptable detection windows. The proposed algorithm is a generalization of Multiple Instance Pruning, which allows to train detectors robust to different image scanning parameters.

machine learning, object detection, object localization, Viola-Jones method, soft cascade, identity document recognition

1. Goode A. Digital identity: solving the problem of trust. Biometric Technology Today. 2019; 2019(10):5–8.

2. Lin WH, Wang P, Tsai CF. Face recognition using support vector model classifier for user authentication. Electronic Commerce Research and Applications. 2016 Jul;18:71–82.

3. Gai K, Qiu M, Sun X. A survey on FinTech. Journal of Network and Computer Applications. 2018 Feb;103:262–73.

4. Bulatov KB, Ingacheva AS, Gilmanov MI, Chukalina MV, Nikolaev DP, Arlazarov VV. Reducing radiation dose for NN-based COVID-19 detection in helical chest CT using real-time monitored reconstruction. Expert Systems with Applications. 2023 Nov;229:120425.

5. Viola P, Jones M. Rapid object detection using a boosted cascade of simple features. In: Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition CVPR 2001. IEEE Comput. Soc;

6. Freund Y, Schapire RE. A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting. Journal of Computer and System Sciences. 1997 Aug;55(1):119–39.

7. Gladilin S., Kotov A., Nikolaev D. and Usilin S. “Postroenie ustoychivykh priznakov detektsii i klassifikatsii obektov, ne obladayuschikh kharakternymi yarkostnymi kontrastami,” ITiVS, no 1, pp. 53-60, 2014.

8. Dollar P, Belongie S, Perona P. The Fastest Pedestrian Detector in the West. In: Procedings of the British Machine Vision Conference 2010. 2010.

9. Li SZ, Zhu L, Zhang Z, Blake A, Zhang H, Shum H. Statistical Learning of Multi-view Face Detection. In: Heyden A, Sparr G, Nielsen M, Johansen P, editors. Computer Vision – ECCV 2002 [Internet]. Berlin, Heidelberg: Springer Berlin Heidelberg; 2002 [cited 2023 Sep 13]. p. 67–81. (Goos G, Hartmanis J, Van Leeuwen J, editors. Lecture Notes in Computer Science; vol. 2353). Available from:

10. Lienhart R, Maydt J. An extended set of Haar-like features for rapid object detection. In: Proceedings International Conference on Image Processing [Internet]. Rochester, NY, USA: IEEE; 2002 [cited 2023 Sep 13]. p. I-900-I–903. Available from:

11. Ramirez GA, Fuentes O. Multi-Pose Face Detection with Asymmetric Haar Features. In: 2008 IEEE Workshop on Applications of Computer Vision [Internet]. Copper Mountain, CO, USA: IEEE; 2008 [cited 2023 Sep 13]. p. 1–6. Available from: http://ieeexplore.ieee.org/document/4544013/

12. Polyakov I., Kuznetsova E., Usilin S. and Nikolaev D. “Training optimal Viola-Jones detectors using greedy algorithms for selecting control parameters with intermediate validation on each level,”Sensory systems, vol. 30, no 3, pp. 241-248, 2016.

13. Chang Huang, Haizhou Ai, Yuan Li, Shihong Lao. Vector boosting for rotation invariant multi-view face detection. In: Tenth IEEE International Conference on Computer Vision (ICCV’05) Volume 1 [Internet]. Beijing, China: IEEE; 2005 [cited 2023 Sep 13]. p. 446-453 Vol. 1. Available from: http://

14. Hutchison D, Kanade T, Kittler J, Kleinberg JM, Mattern F, Mitchell JC, et al. Multi-stage Sampling with Boosting Cascades for Pedestrian Detection in Images and Videos. In: Daniilidis K, Maragos P, Paragios N, editors. Computer Vision– ECCV 2010 [Internet]. Berlin, Heidelberg:

Springer Berlin Heidelberg; 2010 [cited 2023 Sep 13]. p. 196–209. (Lecture Notes in Computer Science; vol. 6316). Available from: http://link.springer.com/10.1007/978-3-642-15567-3_15

16. Bourdev L, Brandt J. Robust object detection via soft cascade. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. 2005.

17. Minkina A, Nikolaev D, Usilin S, Kozyrev V. Generalization of the Viola-Jones method as a decision tree of strong classifiers for real-time object recognition in video stream. In: Seventh International Conference on Machine Vision (ICMV 2014). SPIE; 2015. p. 944517.

19. Wang YQ. An Analysis of the Viola-Jones Face Detection Algorithm. Image Processing On Line. 2014 Jun 26;4:128–48.

20. Shemiakina J, Zhukovsky A, Konovalenko I, Nikolaev D. Automatic cropping of images under projective transformation. In: Nikolaev DP, Radeva P, Verikas A, Zhou J, editors. Eleventh International Conference on Machine Vision (ICMV 2018) [Internet]. Munich, Germany: SPIE; 2019 [cited 2023 Sep 13]. p. 117. Available

21. Usilin SA, Slavin OA, Arlazarov VV. Memory Consumption and Computation Efficiency Improvements of Viola–Jones Object Detection Method for Remote Sensing Applications. Pattern Recognition and Image Analysis. 2021 Jul 1;31(3):571–9.