Adapting Generic Detector for Semi-Supervised Pedestrian Detection

This paper presents a pedestrian detector adaptation approach that aims at adapting a generic detector pre-trained on a public dataset to a specific scene. We hypothesize that the components of the generic detector are useful for constructing a scene-specific detector. For this purpose, our proposed approach is applied to determine the appropriate weights for recombining the components, such that the new detector is expected to perform better in the target scene. Next, the new detector is used to collect new training samples from unlabeled target data. Finally, a weighted CNN is fine-tuned on the extended training set. The experiments demonstrate that the improvements due to the strategies adopted in the proposed approach are significant.

[1]  Meng Wang,et al.  Scene-Specific Pedestrian Detection for Static Video Surveillance , 2014, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[2]  Meng Wang,et al.  Deep Learning of Scene-Specific Classifier for Pedestrian Detection , 2014, ECCV.

[3]  Martial Hebert,et al.  Semi-Supervised Self-Training of Object Detection Models , 2005, 2005 Seventh IEEE Workshops on Applications of Computer Vision (WACV/MOTION'05) - Volume 1.

[4]  Michael S. Bernstein,et al.  ImageNet Large Scale Visual Recognition Challenge , 2014, International Journal of Computer Vision.

[5]  Qingming Huang,et al.  Transferring Boosted Detectors Towards Viewpoint and Scene Adaptiveness , 2011, IEEE Transactions on Image Processing.

[6]  Pietro Perona,et al.  Pedestrian Detection: An Evaluation of the State of the Art , 2012, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[7]  Yuan Shi,et al.  Geodesic flow kernel for unsupervised domain adaptation , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[8]  Luc Van Gool,et al.  Cascaded Confidence Filtering for Improved Tracking-by-Detection , 2010, ECCV.

[9]  Ivor W. Tsang,et al.  Learning Sparse SVM for Feature Selection on Very High Dimensional Datasets , 2010, ICML.

[10]  Bernt Schiele,et al.  Filtered channel features for pedestrian detection , 2015, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[11]  Bernt Schiele,et al.  Filtered Feature Channels for Pedestrian Detection , 2015, CVPR 2015.

[12]  Qiang Yang,et al.  Boosting for transfer learning , 2007, ICML '07.

[13]  Meng Wang,et al.  Automatic adaptation of a generic pedestrian detector to a specific traffic scene , 2011, CVPR 2011.

[14]  Shih-Fu Chang,et al.  Cross-domain learning methods for high-level visual concept classification , 2008, 2008 15th IEEE International Conference on Image Processing.

[15]  Meng Wang,et al.  Transferring a generic pedestrian detector towards specific scenes , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[16]  Fei He,et al.  Cognitive pedestrian detector: Adapting detector to specific scene by transferring attributes , 2015, Neurocomputing.

[17]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[18]  Mubarak Shah,et al.  Online detection and classification of moving objects using progressively improving detectors , 2005, 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05).