Reviewer Info

Thank you for agreeing to review for ICCV 2009. As a reviewer, you are responsible for advising the conference chairs as to the suitability of this paper for presentation at ICCV, and to ensure that the authors receive a fair evaluation of their work. While working on your reviews, please keep in mind the important points below. And of course, remember at all times that you are yourself an author and that you are expecting high quality, detailed, professional reviews of your submitted paper. The other authors are expecting the same from you.

Review content

• Read the paper carefully and in detail.
• Back up claims in your review. If you believe something has been done before, you should say where and by whom. Remarks such as "...has been done before..." without further backing are useless and seriously decrease the credibility of the review. Similarly, if you find something is unclear, you should say what is unclear about it, and provide constructive comments.
• Use you judgment in generating your overall rating. Minor flaws may not be so important. Papers come in many different styles of contributions from totally new, never done before concept, to painstakingly detailed experimental evaluation of existing techniques. Don't use a rigid set of pre-defined criteria. In particular, be careful to properly evaluate applications/systems papers.
• Please explain each recommendation carefully in your review, avoiding empty remarks. In the first box, please list the positive and negative points of the paper carefully.
• Provide detailed reviews. One common complaint from authors is that reviews are sometimes so terse that it is difficult to believe that the reviewer spent sufficient time reading and understanding the paper and it is hard to see how any decision can be made on the basis of such reviews. Given the competitive acceptance rate of ICCV, we owe the authors detailed reviews; short, dismissive are completely useless in the decision process and they are insulting to the authors. We have continued the effort started in other vision conferences, such as CVPR, to reduce the number of papers assigned to reviewers and Area Chairs alike, hoping that this will lead to more thorough review reports.
• You should comment as much as possible on the key areas listed on the review form. At minimum, you need to explain why the paper is or is not appropriate for ICCV and you should establish whether the paper is technical correct and is supported by sound experimentation.
• Remember that your review is for the Area Chair as much as for the authors. Make sure that the AC has sufficiently detailed material to form an opinion. Make sure to use the confidential comments box to communicate with the Area Chair comments that would be inappropriate for the authors, but do not add material that changes substantially the tone of your public comments, since the authors cannot rebut your private comments.
• Avoid personal remarks, and aggressive or inappropriate language. You are reviewing the work, not its authors or the area. In particular general opinions about a field, a research style, or a type of application are not appropriate. Limit yourselves to technical discussions and facts.
• The oral/posters decisions will be made later (at the AC meeting). You should not be concerned with that classification. Your task is only to evaluate whether the paper is sufficiently interesting to recommend publication in ICCV.

Confidentiality

• Do not disclose the content of a paper to others. It is often tempting to casually mention one particular result or approach from a paper that you are currently reviewing to a colleague. In this scenario, in addition to the fact that you have no idea what this colleague's professional relation to the authors might be, you would have effectively publicly released this information without the authors' permission. The safest approach is to never discuss any paper under review and to destroy any material related to the paper after the review is done.
• In the same vein, be careful to not use the results/approaches described in the paper in your own work, even if you plan on crediting the authors eventually. Until the paper is actually published, the authors have not agreed to their ideas being disclosed to and used by others.
• If you feel that one of your students is qualified to review the paper, then it is ok to proceed. However, please do remember that YOU are responsible for the review. This implies that you must check the review and discuss it with the student. The same confidentiality rules applies to the student and it is your responsibility to explain them.

Problem papers

• Conflict of interest: If you recognize the work and you are sure that you are in conflict with the paper, please contact us right away. We (and the software) worked very hard at preventing conflicts but it is impossible to guarantee that a few papers did not slip through. There are many definitions of conflicts of interest (every funding agency has one). Generally, you are in conflict with somebody with whom you currently collaborate closely or have collaborated in the recent past. Beyond rigid definitions, you are definitely in conflict if, having identified the authors, you feel that you cannot produce an objective review.
• Anonymity: There are many gray areas when it comes to the anonymity rules. If you think that a paper reveals the authors through references or other means, please contact us and we will review the situation with the relevant AC. In general, we will tend to lean on the authors' side unless it is obvious that they attempted to influence the review process. There will be no automatic rejection.
• Double submission: Please notify us of any suspicion of double submission to another conference or journal. We will make sure to be flexible and to treat these on a case-by-case basis. There will be no automatic rejection.
• Plagiarism: On the other hand, we will show no flexibility when it comes to plagiarism (which is different from double submission, by the way). Unfortunately, cases of plagiarisms have been steadily increasing, including at major Computer Vision conferences. You need to be on the lookout and remain vigilant.
• Machine-generated papers: A new trend created by people who have too much time on their hands is to submit fake, machine generated papers to conferences. While this was limited in the past to smaller conferences, there is now evidence that such papers are submitted to major conferences. While it is unlikely that such a paper would make it to ICCV, you, as reviewer, are the first line of defense to prevent this kind of abuse of the system.
• Other violations: Other violations such as page limit exceeded or wrong file format are ground for rejection and should be reported immediately.