|
| |
Logbook
|
This page contains information on the different issues that are relevant to CBIR benchmarking.
We centralize ideas here before sorting them out into specific pages.
Our activities correspond to the subtasks we have identified.
-
Image collections: We are gathering images that are free of rights for using them as base for benchmarking.
The collection is available and browsable. See our Data page for details.
-
Image annotation: C. and P. Jorgensen and S. Marchand-Maillet are working in this direction. This work is twofold:
- defining a suitable vocabulary for annotation. This corresponds to the difficult problem of defining suitable ontologies.
- deriving the right tools for facilitating image annotation.
-
Performance measures: A list of performances measures has been derived and proposed.
The idea is to inherit from the experience gathered by the Text Retrieval (TR) community.
For now, we are mostly based on the idea of Precision and Recall. See the following references for more details:
-
Neil J. Gunther, Giordano Beretta.
A Benchmark for Image Retrieval using Distributed Systems over the Internet: BIRDS--I.
HP Labs, Palo Alto, Technical Report HPL--2000--162, San Jose, 2001.
-
Henning Müller, Wolfgang Müller, David McG.\ Squire, Stéphane Marchand--Maillet, Thierry Pun.
Performance Evaluation in Content--Based Image Retrieval: Overview and Proposals,
Pattern Recognition Letters, Vol. 22, No. 5, pp. 593--601, 2001.
|
|
