Life-long Learning with Applications in Monitoring Biodiversity Joachim Denzler Computer Vision Group Michael Stifel Center Jena http://www.inf-cv.uni-jena.de/ http://www.mscj.uni-jena.de/ Friedrich Schiller University Jena Computer Vision Group J. Denzler Life-Long Learning 1
Friedrich Schiller University Jena Outline Computer Vision Group 1 Motivation 2 WALI: Watch, Ask, Learn, and Improve 3 First Results for Biodiversity Research 4 Conclusion J. Denzler Life-Long Learning 2
Friedrich Schiller University Jena Motivation Computer Vision Group Monitoring Biodiversity J. Denzler Life-Long Learning 3
Friedrich Schiller University Jena Motivation Computer Vision Group Automatic Monitoring: AMMOD from Wolfgang W¨ agele: Technical Concept for AMMOD Automated Multi-Sensor Station for Monitoring of Species Diversity (part of BioM-D) J. Denzler Life-Long Learning 4
Friedrich Schiller University Jena Motivation Computer Vision Group Visual Monitoring: Camera Traps J. Denzler Life-Long Learning 5
Friedrich Schiller University Jena Motivation Computer Vision Group Visual Monitoring: Camera Traps J. Denzler Life-Long Learning 6
Friedrich Schiller University Jena Motivation Computer Vision Group Visual Monitoring: Camera Traps J. Denzler Life-Long Learning 7
Friedrich Schiller University Jena Motivation Computer Vision Group Visual Monitoring: Camera Traps J. Denzler Life-Long Learning 8
Friedrich Schiller University Jena Motivation Computer Vision Group Visual Monitoring: Camera Traps J. Denzler Life-Long Learning 9
Friedrich Schiller University Jena Motivation Computer Vision Group Automatic Analysis of Images/Videos Monitoring does not only mean to record data, but also to evaluate it: Million of images per day possible J. Denzler Life-Long Learning 10
Friedrich Schiller University Jena Motivation Computer Vision Group Automatic Analysis of Images/Videos Monitoring does not only mean to record data, but also to evaluate it: Million of images per day possible Presorting of errorneous or useless recordings J. Denzler Life-Long Learning 10
Friedrich Schiller University Jena Motivation Computer Vision Group Automatic Analysis of Images/Videos Monitoring does not only mean to record data, but also to evaluate it: Million of images per day possible Presorting of errorneous or useless recordings Classification of expected species J. Denzler Life-Long Learning 10
Friedrich Schiller University Jena Motivation Computer Vision Group Automatic Analysis of Images/Videos Monitoring does not only mean to record data, but also to evaluate it: Million of images per day possible Presorting of errorneous or useless recordings Classification of expected species Activity detection (eating, sleeping, etc.) J. Denzler Life-Long Learning 10
Friedrich Schiller University Jena Motivation Computer Vision Group Automatic Analysis of Images/Videos Monitoring does not only mean to record data, but also to evaluate it: Million of images per day possible Presorting of errorneous or useless recordings Classification of expected species Activity detection (eating, sleeping, etc.) High accuracy, small error rate J. Denzler Life-Long Learning 10
Friedrich Schiller University Jena Motivation Computer Vision Group Automatic Analysis of Images/Videos Monitoring does not only mean to record data, but also to evaluate it: Million of images per day possible Presorting of errorneous or useless recordings Classification of expected species Activity detection (eating, sleeping, etc.) High accuracy, small error rate Open to changing setups, new species to be detected, etc. J. Denzler Life-Long Learning 10
Friedrich Schiller University Jena Motivation Computer Vision Group Automatic Analysis of Images/Videos Monitoring does not only mean to record data, but also to evaluate it: Million of images per day possible Presorting of errorneous or useless recordings Classification of expected species Activity detection (eating, sleeping, etc.) High accuracy, small error rate Open to changing setups, new species to be detected, etc. Possibility of human to check, correct, and understand the automatic results ( keep the human in the loop ) J. Denzler Life-Long Learning 10
Friedrich Schiller University Jena Motivation Computer Vision Group Automatic Analysis of Images/Videos Monitoring does not only mean to record data, but also to evaluate it: Million of images per day possible Presorting of errorneous or useless recordings Classification of expected species Activity detection (eating, sleeping, etc.) High accuracy, small error rate Open to changing setups, new species to be detected, etc. Possibility of human to check, correct, and understand the automatic results ( keep the human in the loop ) How close are we towards a solution and what are the technical preliminaries? J. Denzler Life-Long Learning 10
Friedrich Schiller University Jena WALI: Watch, Ask, Learn, and Improve Computer Vision Group Outline 1 Motivation 2 WALI: Watch, Ask, Learn, and Improve 3 First Results for Biodiversity Research 4 Conclusion J. Denzler Life-Long Learning 11
Friedrich Schiller University Jena WALI: Watch, Ask, Learn, and Improve Computer Vision Group One Specific Instance: WALI Christoph K¨ ading, Erik Rodner, Alexander Freytag, Joachim Denzler: Watch, Ask, Learn, and Improve: a lifelong learning cycle for visual recognition. European Symposium on Artificial Neural Networks (ESANN). 2016 J. Denzler Life-Long Learning 12
Friedrich Schiller University Jena WALI: Watch, Ask, Learn, and Improve Computer Vision Group One Specific Instance: WALI Key incredients: Multi-class active learning K¨ ading et al.. Active Learning and Discovery of Object Categories in the Presence of Unnameable Instances. IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 4343-4352. 2015. Freytag et al.. Selecting Influential Examples: Active Learning with Expected Model Output Changes European Conference on Computer Vision (ECCV). 562-577. 2014. Novelty detection Bodesheim et al.. Local Novelty Detection in Multi-class Recognition Problems IEEE Winter Conference on Applications of Computer Vision (WACV). 813-820. 2015. Bodesheim et al.. Kernel Null Space Methods for Novelty Detection IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 3374-3381. 2013. Large-scale learning Fr¨ ohlich et al.. Large-Scale Gaussian Process Multi-Class Classification for Semantic Segmentation and Facade Recognition Machine Vision and Applications. 24(5):1043-1053 2013. Rodner et al.. Large-Scale Gaussian Process Classification with Flexible Adaptive Histogram Kernels European Conference on Computer Vision (ECCV). 85-98. 2012. J. Denzler Life-Long Learning 13
Friedrich Schiller University Jena WALI: Watch, Ask, Learn, and Improve Computer Vision Group Multi-class Active Learning Idea: Expected model output change (EMOC) , i.e. ask for a label for that sample that maximizes the change of the model output after knowing the label J. Denzler Life-Long Learning 14
Friedrich Schiller University Jena WALI: Watch, Ask, Learn, and Improve Computer Vision Group Multi-class Active Learning Idea: Expected model output change (EMOC) , i.e. ask for a label for that sample that maximizes the change of the model output after knowing the label p ( y ′ | f ( x ′ )) 1 � � △ f ( x ′ ) = L ( f ( x j ) , f ′ ( x j )) | D | y ′ ∈ Y x j ∈ D We have sample set of labeled and unlabeled samples D , classifier output f ( x j ) given by Gaussian process classifier J. Denzler Life-Long Learning 14
Friedrich Schiller University Jena WALI: Watch, Ask, Learn, and Improve Computer Vision Group Multi-class Active Learning Idea: Expected model output change (EMOC) , i.e. ask for a label for that sample that maximizes the change of the model output after knowing the label p ( y ′ | f ( x ′ )) 1 � � △ f ( x ′ ) = L ( f ( x j ) , f ′ ( x j )) | D | y ′ ∈ Y x j ∈ D We have sample set of labeled and unlabeled samples D , classifier output f ( x j ) given by Gaussian process classifier We need an appropriate loss function: L 1 -loss estimate for the label probability of the given sample: given by the predictive posterior of the Gaussian process classifier used for MC sampling of the labels efficient model update: possible in the Gaussian process framework J. Denzler Life-Long Learning 14
Friedrich Schiller University Jena WALI: Watch, Ask, Learn, and Improve Computer Vision Group WALI: first results (on YouTube videos) Evaluation: number of discovered classes over time, error on the detected classes, error for all classes J. Denzler Life-Long Learning 15
Friedrich Schiller University Jena WALI: Watch, Ask, Learn, and Improve Computer Vision Group WALI: first results (on YouTube videos) Evaluation: Long term behaviour J. Denzler Life-Long Learning 16
Friedrich Schiller University Jena First Results for Biodiversity Research Computer Vision Group Outline 1 Motivation 2 WALI: Watch, Ask, Learn, and Improve 3 First Results for Biodiversity Research 4 Conclusion J. Denzler Life-Long Learning 17
Friedrich Schiller University Jena First Results for Biodiversity Research Computer Vision Group Preliminaries Data : Initially annotated ( large ) data set of species to be detected (standard in DNA-barcoding) J. Denzler Life-Long Learning 18
Recommend
More recommend
