This is the trophy we won at the PASCAL Visual Objects Challenge:

The goal of the The PASCAL Visual Object Classes Challenge is to recognize objects from a number of visual object classes in realistic scenes (i.e. not pre-segmented objects). It is fundamentally a supervised learning problem in that a training set of labelled images was provided. The task was to decide whether an object of a certain class is present in an image or not.

Using the approach described in our CVPR paper and some extensions, we (Thomas Deselaers and Daniel Keysers) participated in this evaluation and obtained quite good results.

T. Deselaers, D. Keysers, and H. Ney. Discriminative Training for Object Recognition using Image Patches. In CVPR 2005, International Conference on Computer Vision and Pattern Recognition, San Diego, CA, June 2005. In press.

Here are the slides of the talk I gave at the PASCAL workshop in April 2005

The detailed results are given in the following tables. More information on the tasks and the submissions of other groups can be obtained from

• The PASCAL website
• The document describing the results

ECR=Equal Classification Rate = 1-EER
AUC=Area Under ROC Curve
for both performance measures higher values are better

Note that in the tables below the results of INRIA are not exactly comparable to the other results because INRIA performed several runs on the test data and only submitted the best runs.

Task 2.1: Classification of motorbikes

ECR	AUC	Group
0.798	0.865	INRIA
0.769	0.829	Aachen
0.767	0.825	Aachen
0.698	0.765	MPI Tuebingen
0.698	0.710	Edinburgh
0.683	0.716	Darmstadt
0.663	0.706	Darmstadt
0.635	0.675	Helsinki
0.624	0.693	Helsinki
0.614	0.666	Helsinki
0.594	0.637	Helsinki

Task 2.2: Classification of bicycles

ECR	AUC	Group
0.728	0.813	INRIA
0.667	0.724	Aachen
0.665	0.729	Aachen
0.616	0.654	MPI Tuebingen
0.616	0.645	Helsinki
0.604	0.647	Helsinki
0.575	0.606	Edinburgh
0.527	0.567	Helsinki
0.524	0.546	Helsinki

Task 2.3: Classification of people

ECR	AUC	Group
0.719	0.798	INRIA
0.669	0.739	Aachen
0.663	0.721	Aachen
0.614	0.661	Helsinki
0.601	0.650	Helsinki
0.591	0.655	MPI Tuebingen
0.587	0.630	Helsinki
0.574	0.618	Helsinki
0.519	0.552	Edinburgh

Task 2.4: Classification of cars

ECR	AUC	Group
0.720	0.802	INRIA
0.716	0.780	Aachen
0.703	0.767	Aachen
0.692	0.744	Helsinki
0.677	0.717	MPI Tuebingen
0.676	0.740	Helsinki
0.655	0.709	Helsinki
0.644	0.694	Helsinki
0.633	0.655	Edinburgh
0.551	0.572	Darmstadt

Task 1.1: Classification of motorbikes

ECR	AUC	Group
0.977	0.998	INRIA
0.972	0.994	Southampton
0.968	0.997	INRIA
0.964	0.996	INRIA
0.949	0.989	Southampton
0.940	0.987	Aachen
0.940	0.985	Southampton
0.926	0.979	Aachen
0.921	0.974	Helsinki
0.917	0.970	Helsinki
0.912	0.952	Helsinki
0.903	0.966	Ankara
0.898	0.960	Helsinki
0.875	0.945	MPI Tuebingen
0.856	0.882	Darmstadt
0.829	0.919	Darmstadt
0.722	0.765	Edinburgh

Task 1.2: Classification of bicycles

ECR	AUC	Group
0.930	0.982	INRIA
0.930	0.981	INRIA
0.918	0.974	INRIA
0.895	0.961	Southampton
0.868	0.954	Aachen
0.868	0.943	Southampton
0.851	0.930	Southampton
0.842	0.931	Aachen
0.816	0.895	Helsinki
0.795	0.891	Helsinki
0.781	0.864	Helsinki
0.781	0.822	Ankara
0.767	0.880	Helsinki
0.754	0.838	MPI Tuebingen
0.689	0.724	Edinburgh

Task 1.3: Classification of people

ECR	AUC	Group
0.917	0.979	INRIA
0.917	0.972	INRIA
0.901	0.965	INRIA
0.881	0.943	Southampton
0.861	0.936	Aachen
0.861	0.928	Aachen
0.857	0.921	Helsinki
0.850	0.927	Helsinki
0.845	0.919	Helsinki
0.841	0.925	Southampton
0.833	0.931	Helsinki
0.833	0.918	Southampton
0.803	0.816	Ankara
0.731	0.834	MPI Tuebingen
0.571	0.597	Edinburgh

Task 1.4: Classification of cars

ECR	AUC	Group
0.961	0.992	INRIA
0.938	0.987	INRIA
0.937	0.983	INRIA
0.925	0.978	Aachen
0.920	0.979	Aachen
0.913	0.972	Southampton
0.909	0.971	Helsinki
0.908	0.968	Helsinki
0.901	0.961	Southampton
0.898	0.959	Southampton
0.869	0.956	Helsinki
0.847	0.934	Helsinki
0.840	0.920	Ankara
0.831	0.918	MPI Tuebingen
0.793	0.798	Edinburgh
0.644	0.717	Darmstadt
0.548	0.578	Darmstadt

;