We consider a semi-supervised approach to the problem of track classification in dense three-dimensional range data. This problem involves the classification of objects that have been segmented and tracked without the use of a class-specific tracker. This paper is an extended version of our previous work.
We propose a method based on the expectation–maximization algorithm: iteratively (1) train a classifier, and (2) extract useful training examples from unlabeled data by exploiting tracking information. We evaluate our method on a large multiclass problem in dense range data collected from natural street scenes.
When given only three hand-labeled training tracks of each object class, the final accuracy of the semi-supervised algorithm is comparable to that of the fully supervised equivalent which uses two orders of magnitude more. Further, we show experimentally that the accuracy of a classifier considered as a function of human labeling effort can be substantially improved using this method. Finally, we show that a simple algorithmic speedup based on incrementally updating a boosting classifier can reduce learning time by a factor of three.
AliKHaslerDFleuretF (2011) FlowBoost – Appearance learning from sparsely annotated video. In: 24th IEEE conference on computer vision and pattern recognition, Colorado Springs, USA, 21–23 June 2011,1433–1440. Piscataway: IEEE Press.
2.
BlumAMitchellT (1998) Combining labeled and unlabeled data with co-training. In: 11th annual conference on computational learning theory (COLT ’98), Madison, USA, 24–26 July 1998,92–100. New York: ACM.
3.
ChristoudiasCMUrtasunRKapoorADarrellT (2009) Co-training with noisy perceptual observations. In: 22nd IEEE conference on computer vision and pattern recognition, Miami, USA, 20–25 June 2009, pp. 2844–2851. Piscataway: IEEE Press.
4.
DahlkampHKaehlerAStavensDThrunSBradskiG (2006) Self-supervised monocular road detection in desert terrain. In: Robotics: science and systems II (ed SukhatmeGSSchaalSBurgardWFoxD), Philadelphia, USA, 16–19 August 2006. Cambridge: MIT Press.
5.
DalalNTriggsB (2005) Histograms of oriented gradients for human detection. In: 18th IEEE conference on computer vision and pattern recognition, San Diego, USA, 20–26 June 2005, vol. 2, pp. 886–893. Piscataway: IEEE Press.
6.
DouillardBFoxDRamosFDurrant-WhyteH (2011) Classification and semantic mapping of urban environments. International Journal of Robotics Research30(1): 5–32.
7.
FriedmanJHastieTTibshiraniR (2000) Additive logistic regression: a statistical view of boosting. Annals of Statistics28(2): 337374.
8.
GodecMRothPBischofH (2011) Hough-based tracking of non-rigid objects. In: 13th international conference on computer vision, Barcelona, Spain, 6–13 November 2011, pp. 81–88. Piscataway: IEEE Press.
9.
JohnsonAHebertM (1999) Using spin images for efficient object recognition in cluttered 3D scenes. IEEE Transactions on Pattern Analysis and Machine Intelligence21(5): 433–449.
10.
KalalZMatasJMikolajczykK (2010) P-N learning: bootstrapping binary classifiers by structural constraints. In: IEEE conference on computer vision and pattern recognition, San Francisco, USA, 13–18 June 2010, pp. 49–56. Piscataway: IEEE Press.
11.
LaiKFoxD (2009) 3D Laser scan classification using web data and domain adaptation. In: Robotics: science and systems IV (ed TrinkleJMatsuokaYCastellanosJA), Seattle, USA, 28 June – 1 July 2009. Cambridge: MIT Press.
12.
LaiKFoxD (2010) Object recognition in 3D point clouds using web data and domain adaptation. International Journal of Robotics Research29(8): 1019–1037.
13.
LevinsonJThrunS (2010) Unsupervised calibration for multi-beam lasers. In: 12th international symposium on experimental robotics, Delhi, India, 18–21 December 2010.
14.
LiuWHeJChangS-F (2010) Large graph construction for scalable semi-supervised learning. In: 27th international conference on machine learning (ICML 2010), Haifa, Israel, 21–24 June 2010.
15.
LookingbillARogersJLiebDCurryJThrunS (2007) Reverse optical flow for self-supervised adaptive autonomous robot navigation. International Journal of Computer Vision74(3): 287–302.
16.
LuberMArrasKOPlagemannCBurgardW (2008) Classifying dynamic objects: an unsupervised learning approach. In: Robotics: science and systems IV (ed BrockOTrinkleJRamosF), Zurich, Switzerland, 25–28 June 2008. Cambridge: MIT Press.
17.
MontemerloMBeckerJBhatSDahlkampHDolgovDEttingerS. (2008) Junior: the Stanford entry in the urban challenge. Journal of Field Robotics25(9): 570–597.
18.
PanSJYangQ (2010) A survey on transfer learning. IEEE Transactions on Knowledge and Data Engineering22(10): 1345–1359.
19.
SocherRFei-FeiL (2010) Connecting modalities: semi-supervised segmentation and annotation of images using unaligned text corpora. In: IEEE conference on computer vision and pattern recognition, San Francisco, USA, 13–18 June 2010, pp. 966–973. Piscataway: IEEE Press.
20.
SpinelloLTriebelRSiegwartR (2010) Multiclass multimodal detection and tracking in urban environments. International Journal of Robotics Research29(12): 1–35.
21.
StalderSGrabnerHGoolLV (2009) Beyond semi-supervised tracking: tracking should be as simple as detection, but not simpler than recognition. In: 3rd on-line learning for computer vision workshop (OLCV ’09), Kyoto, Japan, 3 October 2009.
22.
TangFBrennanSZhaoQTaoH (2007) Co-tracking using semi-supervised support vector machines. In: 11th international conference on computer vision (ICCV 2007), Santa Cruz, USA, 14–21 October 2007, pp. 1–8. Piscataway: IEEE Press.
23.
TeichmanALevinsonJThrunS (2011a) Towards 3D object recognition via classification of arbitrary object tracks. In: international conference on robotics and automation (ICRA 2011), Shanghai, China, 9–13 May 2011, pp. 4034–4041. Piscataway: IEEE Press.
TorralbaAMurphyKFreemanW (2004) Sharing features: efficient boosting procedures for multiclass object detection. In: IEEE conference on computer vision and pattern recognition, Washington DC, USA, 27 June – 2 July 2004, vol. 2, pp. 762–769. Piscataway: IEEE Press.
26.
TriebelRShinJSiegwartR (2010) Segmentation and unsupervised part-based discovery of repetitive objects. In: Robotics: science and systems II (ed MatsuokaYDurrant-WhyteHNeiraJ), Zaragoza, Spain, 27–30 June 2010. Cambridge: MIT Press.
27.
WurmKMKummerleRStachnissCBurgardW (2009) Improving robot navigation in structured outdoor environments by identifying vegetation from laser data. In: 2009 IEEE/RSJ international conference on intelligent robots and systems, St. Louis, USA, 11–15 October 2009, pp. 1217–1222. Piscataway: IEEE Press.
ZhuXGhahramaniZLaffertyJ (2003) Semi-supervised learning using Gaussian fields and harmonic functions. In: 20th international conference on machine learning (ICML 2003), Washington DC, USA, 21–24 August 2004.
