this is for holding javascript data
Dario WurmD edited section_Reviewer_Comments_for_Authors__.tex
over 8 years ago
Commit id: 1da7b15a2bcd2596606a75c60989ef0295a22e8d
deletions | additions
diff --git a/section_Reviewer_Comments_for_Authors__.tex b/section_Reviewer_Comments_for_Authors__.tex
index f32191c..721e862 100644
--- a/section_Reviewer_Comments_for_Authors__.tex
+++ b/section_Reviewer_Comments_for_Authors__.tex
...
2. The details of the main technical components of this work are not provided. As for false positive class, I don’t see how it is used in the re-id system.
\end{quote}
\textit{We thank the reviewer for pointing this out. If this matter is not clear, we shall endeavor to \textbf{make} it clear.}
\textit{In figure 2, we assume the Single-frame Classifier is a classifier that will classify input images as one of the persons available in the training set.
Therefor, Therefore, when a classifier that only has classes for people, receives an input image of a fire-extinguisher (figure 5) it always fails to correctly classify (there is no correct class!).}
\textit{In Section 5.2.2 Single frame Re-Identification we cite [19] as the algorithm employed to enact classification. [19] describes a multi-class classifier, that does not, by default, contains an extra class for unclassifiable inputs. Therefore, when using the "FP Class module", we are in fact providing the re-identification classifier algorithm with extra training samples for it to train an extra class, a class of detection false positives.}
\textit{For this to be clearer we have edited the 5.4 Scenarios section thusly:}
...
\begin{quote}
6. I don’t see the difference between the original precision/recall and the proposed precision/recall. I think the paper needs more specific description on them.
\end{quote}
\textit{Well pointed out. We did not mean to imply we had innovated in this aspect,
therefor therefore we edited the corresponding part of the abstract to}
\textbf{\textit{In this paper, we apply precision and recall metrics to evaluate such integrated system}}
...
3. I think the Window-based Classifier is exactly multi-shot. Some other multi-shot methods also do not require contiguous frames and in-camera trackers. The proposed Window-based Classifier is straightforward and reasonable. As an important contribution, it's better to compare it with other other multi-shot methods, such as reference [4].
\end{quote}
\textit{We thank the reviewer to point this out, as it makes clear that we must clarify in the text that multi-shot algorithms that we are aware of (such as [4]) require a way to specify which images belong to a single person (such as a tracker or manual selection) prior to being able to analyze those images in a multi-shot fashion. No multi-shot algorithm that we are aware of does some kind of unsupervised clustering first to try and discover which images belong to a same individual, to then be able to use such data in a multi-shot fashion without a tracker or human intervention.
Therefor, Therefore, we have enhanced the clarifying paragraph thusly:}
\textbf{\textit{Although this procedure is similar to multi-shot, it requires less information. In this work, window-based classification works with any single-shot re-identification algorithm, and does not require an in-camera tracker or manual selection of the images that belong to each individual, contrary to all of the works that do multi-shot re-identification known to the authors.}}