Learning to detect partially overlapping instances

Carlos Arteta, Victor Lempitsky, J. Alison Noble, Andrew Zisserman

    Research output: Contribution to journalConference articlepeer-review

    37 Citations (Scopus)


    The objective of this work is to detect all instances of a class (such as cells or people) in an image. The instances may be partially overlapping and clustered, and hence quite challenging for traditional detectors, which aim at localizing individual instances. Our approach is to propose a set of candidate regions, and then select regions based on optimizing a global classification score, subject to the constraint that the selected regions are non-overlapping. Our novel contribution is to extend standard object detection by introducing separate classes for tuples of objects into the detection process. For example, our detector can pick a region containing two or three object instances, while assigning such region an appropriate label. We show that this formulation can be learned within the structured output SVM framework, and that the inference in such model can be accomplished using dynamic programming on a tree structured region graph. Furthermore, the learning only requires weak annotations - a dot on each instance. The improvement resulting from the addition of the capability to detect tuples of objects is demonstrated on quite disparate data sets: fluorescence microscopy images and UCSD pedestrians.

    Original languageEnglish
    Article number6619259
    Pages (from-to)3230-3237
    Number of pages8
    JournalProceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
    Publication statusPublished - 2013
    Event26th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2013 - Portland, OR, United States
    Duration: 23 Jun 201328 Jun 2013


    Dive into the research topics of 'Learning to detect partially overlapping instances'. Together they form a unique fingerprint.

    Cite this