INRIA Horses

Vittorio Ferrari, Frederic Jurie, Cordelia Schmid


Dataset for testing object class detection algorithms.

It contains 170 images containing one or more horses horses collected from the Internet (positive images), and 170 images without horses (negative images).

All horses in all images are annotated with a bounding-box. The main challenges it offers are clutter, intra-class shape variability, and scale changes. The horses are mostly unoccluded, taken from approximately the side viewpoint, and face the same direction.

The package also includes edgemaps produced by the excellent Berkeley natural boundary detector.

This dataset appeared in the CVPR 2007 [1], PAMI 2008 [2], and IJCV 2010 [3] papers.


Filename Description Size
inria-horses-v103.tgz Object annotations for all horses. We also include edgemaps produced by the excellent Berkeley ‘natural boundary detector’. 44MB
README.txt Description of contents 3KB


New in V 1.02 and 1.03:

  • adjusted ground-truth of image 194 to fit the image size
  • fixed ground-truth annotation for images 4047 and 4111.


  1. Accurate Object Detection with Deformable Shape Models Learnt from Images
    V. Ferrari, F. Jurie, and C. Schmid,
    In Computer Vision and Pattern Recognition (CVPR), 2007.
  2. Groups of Adjacent Contour Segments for Object Detection
    V. Ferrari, L. Fevrier, F. Jurie, and C. Schmid,
    In IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 2008.
  3. From Images to Shape Models for Object Detection
    V. Ferrari, F. Jurie, and C. Schmid,
    In International Journal of Computer Vision (IJCV), 2010.