In order for computer vision to mature from both scientific and industrial points of view, we must have objective methods for evaluating algorithms. Towards this end, the purpose of this workshop is to bring together researchers in the computer vision community to present papers and discuss methods for empirical evaluation of vision algorithms. The focus of the workshop is on experimental methods. The workshop will include invited speakers, panel discussions, and presentations of acepted papers. We hope to foster interaction and discussion on this subject, and to learn from related fields that have sucessfully used evaluation methods.

Click here to see post-workshop summary.

Click here to see the advance program of the workshop!

The workshop includes invited speakers and presentations of contributed papers. In addition, editors of some of the major vision journals have agreed to participate in a panel discussion on the theme "Expectations for papers on empirical evaluation submitted to computer vision journals."
scope of the workshop includes but is not limited to

Contributed papers will be reviewed based on content relevant to the theme of the workshop, with an emphasis on sound empirical methods and results. Relevant topics may include, but are not limited to:

• database collection,
• independent evaluation methods,
• comparison of different methods to solve classical computer vision problems,
• evaluation of medical imaging algorithms in clinical settings,
• evaluation of algorithms for industrial applications,
• and design of experiments.

3 copies of papers should be received no later than 30 January, 1998 to:

Kevin Bowyer
Computer Science & Engineering
University of South Florida
4302 East Fowler Avenue, ENB 326
Tampa, FL 33620-5399

Each submission should have a cover page which includes the address, phone number, fax number, and e-mail address of the corresponding author. Papers are limited to 30 double-spaced pages (12 pt, 1 inch margins), including fiqures, references, etc. The papers will be reviewed based on content relevent to the scope of the workshop with an emphasis on empirical results. Accepted papers will be published as part of an edited book.

Click here for European Computer Vision Net Benchmarking web page.

Click here for link to Workshop on Evaluation and Validation of CV Algorithms

U of Toronto's "Delve" - "Data for Evaluating Learning in Valid Experiments"