Learning to segment from a few well-selected training images

We address the task of actively learning a segmentation system: given a large number of unsegmented images, and access to an oracle that can segment a given image, decide which images to provide, to quickly produce a segmenter (here, a discriminative random field) that is accurate over this distribution of images. We extend the standard models for active learner to define a system for this task that first selects the image whose expected label will reduce the uncertainty of the other unlabeled images the most, and then after greedily selects, from the pool of unsegmented images, the most informative image. The results of our experiments, over two real-world datasets (segmenting brain tumors within magnetic resonance images; and segmenting the sky in real images) show that training on very few informative images (here, as few as 2) can produce a segmenter that is as good as training on the entire dataset.

[1]  J. Besag.  On the Statistical Analysis of Dirty Pictures , 1986 .

[2]  David A. Cohn,et al.  Active Learning with Statistical Models , 1996, NIPS.

[3]  William A. Gale,et al.  A sequential algorithm for training text classifiers , 1994, SIGIR '94.

[4]  David D. Lewis,et al.  A sequential algorithm for training text classifiers: corrigendum and additional data , 1995, SIGF.

[5]  Nello Cristianini,et al.  Query Learning with Large Margin Classifiers , 2000, ICML.

[6]  Greg Schohn,et al.  Less is More: Active Learning with Support Vector Machines , 2000, ICML.

[7]  Andrew McCallum,et al.  Toward Optimal Active Learning through Sampling Estimation of Error Reduction , 2001, ICML.

[8]  Daphne Koller,et al.  Support Vector Machine Active Learning with Applications to Text Classification , 2002, J. Mach. Learn. Res..

[9]  Tsuhan Chen,et al.  An active learning framework for content-based information retrieval , 2002, IEEE Trans. Multim..

[10]  Andrew Zisserman,et al.  Classifying Images of Materials: Achieving Viewpoint and Illumination Independence , 2002, ECCV.

[11]  Xiaowei Xu,et al.  Representative Sampling for Text Classification Using Support Vector Machines , 2003, ECIR.

[12]  Martial Hebert,et al.  Discriminative random fields: a discriminative framework for contextual interaction in classification , 2003, Proceedings Ninth IEEE International Conference on Computer Vision.

[13]  H. Sebastian Seung,et al.  Selective Sampling Using the Query by Committee Algorithm , 1997, Machine Learning.

[14]  Arnold W. M. Smeulders,et al.  Active learning using pre-clustering , 2004, ICML.

[15]  Pietro Perona,et al.  Learning Generative Visual Models from Few Training Examples: An Incremental Bayesian Approach Tested on 101 Object Categories , 2004, 2004 Conference on Computer Vision and Pattern Recognition Workshop.

[16]  Pietro Perona,et al.  Learning Generative Visual Models from Few Training Examples: An Incremental Bayesian Approach Tested on 101 Object Categories , 2004, 2004 Conference on Computer Vision and Pattern Recognition Workshop.

[17]  Russell Greiner,et al.  Optimistic Active-Learning Using Mutual Information , 2007, IJCAI.

[18]  Jaime G. Carbonell,et al.  Optimizing estimated loss reduction for active sampling in rank learning , 2008, ICML '08.

[19]  Chi-Hoon Lee,et al.  Segmenting Brain Tumors Using Pseudo-Conditional Random Fields , 2008, MICCAI.

[20]  Kristen Grauman,et al.  Multi-Level Active Prediction of Useful Image Annotations for Recognition , 2008, NIPS.

[21]  Fei-Fei Li,et al.  Towards Scalable Dataset Construction: An Active Learning Approach , 2008, ECCV.

[22]  Derek Hoiem,et al.  Learning CRFs Using Graph Cuts , 2008, ECCV.