diff options
| -rw-r--r-- | report/paper.md | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/report/paper.md b/report/paper.md index 46dd82e..f9fb19a 100644 --- a/report/paper.md +++ b/report/paper.md @@ -251,7 +251,7 @@ The difference between the top $k$ accuracies of the two methods gets smaller as \end{center} \end{figure} -Figure \ref{fig:rerank} shows successful reranking as performed using the query and gallery sets. The $k$-reciprocal ranklist of the second nearest neighbour, contains the target identity, and when the final distance is calculated as a function original and Jaccard distance, it becomes the closest neighbour. This happens with three cases of the identity, achieving 100% mAP despite the image visually hard to classify due to the presence of another person in the photos and a completely different pose. +Figure \ref{fig:rerank} shows successful reranking as performed using the query and gallery sets. The $k$-reciprocal ranklist of the second nearest neighbour, contains the target identity, and when the final distance is calculated as a function original and Jaccard distance, it becomes the closest neighbour. This happens with three cases of the identity, despite the image visually hard to classify due to the presence of another person in the photos and a completely different pose. \begin{figure} \begin{center} @@ -281,7 +281,7 @@ Re-ranking with Jaccard & \textit{O(N\textsuperscript{2}logN)} \\ \hl Overall the re-ranking method gives a significant improvement to both top $k$ accuracy and mean average precision. The cost of this operation is an increase in computation time due to the change in complexity from -the baseline, summarized in table \ref{tab:complexity}. +the baseline, summarized in the table. # References |