TR2010-037

Breaking the Interactive Bottleneck in Multi-Class Classification with Active Selection and Binary Feedback


    •  Joshi, A.J., Porikli, F.M., Papanikolopoulos, N., "Breaking the Interactive Bottleneck in Multi-class Classification with Active Selection and Binary Feedback", IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2010.
      BibTeX TR2010-037 PDF
      • @inproceedings{Joshi2010jun,
      • author = {Joshi, A.J. and Porikli, F.M. and Papanikolopoulos, N.},
      • title = {Breaking the Interactive Bottleneck in Multi-class Classification with Active Selection and Binary Feedback},
      • booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
      • year = 2010,
      • month = jun,
      • url = {https://www.merl.com/publications/TR2010-037}
      • }
  • Research Areas:

    Artificial Intelligence, Computer Vision, Machine Learning

Abstract:

Multi-class classification schemes typically require human input in the form of precise category names or numbers for each example to be annotated - providing this can be impractical for the user when a large (and possibly unknown) number of categories are present. In this paper, we propose a multi-class active learning model that requires only binary (yes/no type) feedback from the user. For instance, given two images the user only has to say whether they belong to the same class or not. We first show the interactive benefits of such a scheme with experiments. We then propose a Value of Information (VOI)-based active selection algorithm in the binary feedback model. The algorithm iteratively selects image pairs for annotation so as to maximize accuracy, while also minimizing user annotation effort. To our knowledge, this is the first multi-class active learning approach that requires only yes/no inputs. Experiments show that the proposed method can substantially minimize user supervision compared to the traditional training model, on problems with as many as 100 classes. We also demonstrate that the system is robust to real-world issues such as class population imbalance and labeling noise.

 

  • Related News & Events

    •  NEWS    CVPR 2010: 8 publications by C. Oncel Tuzel, Tim K. Marks, Yuichi Taguchi, Srikumar Ramalingam, Michael J. Jones and Amit K. Agrawal
      Date: June 13, 2010
      Where: IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
      MERL Contacts: Michael J. Jones; Tim K. Marks
      Brief
      • The papers "Optimal Coded Sampling for Temporal Super-Resolution" by Agrawal, A.K., Gupta, M., Veeraraghavan, A.N. and Narasimhan, S.G., "Breaking the Interactive Bottleneck in Multi-class Classification with Active Selection and Binary Feedback" by Joshi, A.J., Porikli, F.M. and Papanikolopoulos, N., "Axial Light Field for Curved Mirrors: Reflect Your Perspective, Widen Your View" by Taguchi, Y., Agrawal, A.K., Ramalingam, S. and Veeraraghavan, A.N., "Morphable Reflectance Fields for Enhancing Face Recognition" by Kumar, R., Jones, M.J. and Marks, T.K., "Increasing Depth Resolution of Electron Microscopy of Neural Circuits using Sparse Tomographic Reconstruction" by Veeraraghavan, A., Genkin, A.V., Vitaladevuni, S., Scheffer, L., Xu, S., Hess, H., Fetter, R., Cantoni, M., Knott, G. and Chklovskii, D., "Specular Surface Reconstruction from Sparse Reflection Correspondences" by Sankaranarayanan, A., Veeraraghavan, A.N., Tuzel, C.O. and Agrawal, A.K., "Fast Directional Chamfer Matching" by Liu, M.-Y., Tuzel, C.O., Veeraraghavan, A.N. and Chellappa, R. and "Robust RVM regression using sparse outlier model" by Mitra, K., Veeraraghavan, A. and Chellappa, R. were presented at the IEEE Conference on Computer Vision and Pattern Recognition (CVPR).
    •