calvin upper-body detector v1.04

Marcin Eichner, Vittorio Ferrari


We release here software for human upper body detection in still images. It is based on the successful part-based object detection framework [4] and contains a model to detect near-frontal upper-bodies, trained from the data of [3]. The resulting detector returns bounding-boxes fitting the head and upper half of the torso of the person.

In order to find more people we complement the primary detector with upper-body detections regressed from the Viola-Jones [5] face detector. This is especially valuable for people in poses difficult to detect by the upper-body model (e.g. arms raised above the head). Both primary and secondary detectors are combined by this release and a single homogeneous set of upper-body bounding-boxes are returned.

The bounding-boxes returned by the detector released here can be directly fed into our pose estimation software [1], which includes a matlab routine to easily interface with this detector. By installing both this detector and [1] you get a complete and fully automatic human detection and pose estimation pipeline.

This upper-body detector improves over [3] in that:
  • it is easily portable to any Linux platform, as it is based on the excellent code of [4]
  • the primary upper-body detector based on [4] performs better than the one in [3] based on [7]
  • more persons are found thanks to the complementary face detector
  • new in v1.04:


    this detector has been evaluated in our Technical Report [8]

    Training data

    The upper-body detector was trained from the data of [3].


    calvin_upperbody_detector_v1.04.tgz calvin upper-body detector 223 kB
    README.html description of contents 31 kB
    voc-release3.1.tgz snapshot of the object dectection framework [4] required by our upper-body detector 7200 kB


    [1] Eichner, M. and Ferrari, V.
    2d articulated human pose estimation code

    [2] Eichner, M. and Ferrari, V.
    Better Appearance Models for Pictorial Structures
    Proceedings of British Machine Vision Conference (BMVC), 2009.
    Document: PDF

    [3] M. Marin, V. Ferrari, A. Zisserman
    upper-body detector

    [4] P. Felzenszwalb, R. Girshick, D. McAllester, D. Ramanan
    Object Detection with Discriminatively Trained Part Based Models
    Pattern Recognition and Machine Learning (PAMI), 2009

    [5] P. Viola, M. Jones
    Rapid Object Detection using a Boosted Cascade of Simple Features
    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2001

    [6] OpenCV computer vision library

    [7] N. Dalal and B. Triggs
    Histograms of Oriented Gradients for Human Detection
    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2005

    [8] M.Eichner, M. Marin-Jimenez, A. Zisserman, V.Ferrari
    Articulated Human Pose Estimation and Search in (Almost) Unconstrained Still Images
    ETH Zurich, D-ITET, BIWI, Technical Report No.272, September 2010.
    Document: PDF


    We thank Pedro Felzenszwalb, David McAllister and Deva Ramanan for allowing us to host their object detection framework [4] on our website.

    We thank Manuel Marin for the training data released on [3].

    We thank Alessandro Prest for his contribution into the training of the part-based upper-body model

    Finally we thank Pietro Perona for challenging us with images of his group during a talk at Caltech. This prompted us to improve the portability and performance of the detector.

    Please report problems with this page to Marcin Eichner
    Last updated on Tuesday, 02nd February, 2016