Category Independent Object Detection
Implementation of Object Detection algorithm:
Matlab implementation of the object detection algorithm [1] (2011-11-04, ver 1.1)
Notes:
The algorithm needs the executables of Pedro F. Felzenszwalb's and Daniel P. Huttenlocher's superpixel segmentation algorithm. You can download the code from: Felzenszwalb's page.
- The code includes the SS feature developed and described [2]. If you use results derived from this code in your publications, please cite [1], [2], and [4].
Precomputed windows for VOC2007 dataset:
VOC 2007 boxes (Obtained using VOC training data)
VOC 2007 boxes (Obtained using training data from [2])
Performance evaluation:
Recall-overlap curves, computed using methods described in [1], [2], and [3].
References:
[1] Rahtu E. & Kannala J. & Blaschko M. B. (2011) Learning a Category Independent Object Detection Cascade. Proc. International Conference on Computer Vision (ICCV 2011). (Full paper)
[2] Alexe B. & Deselaers T. & Ferrari V. (2010) What is an object? Proc The IEEE Conference on Computer Vision and Pattern Recognition (CVPR).
[3] Alexe B. & Deselaers T. & Ferrari V. (2011) Measuring the objectness of image windows. ETH Zurich tech report #276. Project page
[4] Felzenszwalb P. & Huttenlocher D. (2004) Efficient Graph-Based Image Segmentation. International Journal of Computer Vision, Volume 59, Number 2.
If you encounter problems or find bugs in these implementations, please contact Esa Rahtu (erahtu at ee.oulu.fi).