Projet ImagEval / ImagEval Project

Evaluation campaign funded by the French "Techno-Vision" program. The ImagEVAL project, funded by the French Techno-Vision program, relates to the evaluation of technologies of image filtering, content-based image retrieval (CBIR) and automatic description of images in large-scale image databases The objective of ImagEVAL is double: * to organize important evaluations starting from concrete needs * to evaluate technologies held by national and foreign research laboratories, and software solutions.

Different tasks have been defined by ImagEVAL committee (detailled tasks description): ImagEVAL a organisé l'évaluation d'algorithmes pour les cinq tâches suivantes:

• Recognition of transformed images
• Text/Image Mixed research
• Detection of text areas
• Detection of objects
• Semantics extraction

The CMM proposed a connected component (CC) approach for text localization (task 3) which was ranked first in the evaluation campaign. The official database contains 500 images including old postcards, stamps, indoor and outdoor scene images and also images without any textual data. Furthermore images can be grayscale or color. More precisely, the database contains: 190 old postcards (mainly with legends), 206 color photographs (mainly text within the scene),104 black and white photographs (mainly text within the scene), 90 images without any text area: 40 old postcards, 47 color photographs, 3 black and white photographs. In spite of the wide variety of text and images, a promising score is obtained for the whole official database. 60% text areas were correctly detected. Official results.

	Precision	Recall	f-mean
ICDAR metric	0.49	0.69	0.57
Wolf&Jolion metric	0.54	0.69	0.60

Some results in images:

For further information on this work:

• PhD of Thomas Retornaz (in french).
• ISMM 2007.
• Workshop ImagEval.