diff --git a/presentation.tex b/presentation.tex index 7e1d5c5..dba9358 100644 --- a/presentation.tex +++ b/presentation.tex @@ -56,7 +56,7 @@ \newacronym{acr::lod}{LoD}{Level of Detail} -\newacronym{acr::elod}{eLoD}{evalution Level of Detail} +\newacronym{acr::elod}{eLoD}{evaluation Level of Detail} \newacronym{acr::lidar}{LiDAR}{Light Detection and Ranging} \newacronym{acr::dsm}{DSM}{Digital Surface Model} \newacronym{acr::gui}{GUI}{Graphical User Interface} @@ -90,20 +90,20 @@ \only<1-4>{ \begin{itemize}[label=$\blacktriangleright$, font=\color{IGNGreen}, itemsep=2em] \item<1-> Automatic urban modeling: an active research area~\citep{Musialski2012}; - \item<2-> Results seamless but lack generality~\citep{rottensteiner2014results}; + \item<2-> Results seamless but lack generality and often erroneous~\citep{rottensteiner2014results}; \begin{itemize}[label=$\longrightarrow$] \item<3-> labourious manual corrections. \end{itemize} - \item<4-> Urban 3D model semantic diagnostic not well studied~\citep{nguatem2017modeling}; + \item<4-> Urban 3D model semantic diagnostic \textcolor{purple}{not well studied}~\citep{nguatem2017modeling}; \end{itemize} } \only<5->{ \begin{itemize}[label=Goal $\longrightarrow$, font=\color{purple}, leftmargin=2cm] - \item<5-> Detect and describe semantic errors that affects building 3D models. + \item<5-> Detect and describe semantic errors that affect building 3D models. \end{itemize} \begin{itemize}[label=$\blacktriangleright$, font=\color{IGNGreen}, itemsep=2em] \item<6-> Semantic errors independent from \textbf{reconstruction methods} and \textbf{urban scenes}. - \item<7-> Transferablility, and hence scallability, of the evalution method. + \item<7-> \textbf{Transferability}, and hence scallability, of the evaluation method. \end{itemize} } \end{frame} @@ -120,15 +120,15 @@ \begin{frame}{Main ideas behind our approach} \begin{itemize}[label=$\blacktriangleright$, font=\color{IGNGreen}, itemsep=2em] \item<1-> Compile errors that affect building models in a taxonomy; - \item<2-> Evaluation at building level $\Longrightarrow$ formulated as a supervized classification problem; - \item<3-> Study in a \textbf{2.5D overhead} modeling setting. + \item<2-> Evaluation at building level $\Longrightarrow$ formulated as a supervised classification problem; + \item<3-> Study in a \textbf{2.5D overhead (aerial)} modeling setting. \end{itemize} \end{frame} \begin{frame}{Taxonomy structure} Two criteria determine the taxonomy structure: \begin{itemize}[label=$\blacktriangleright$, font=\color{IGNGreen}, itemsep=2em] - \item<2-> the \acrshort{acr::lod}; - \item<3-> the \emph{finesse}: the semantic evaluation level. + \item<2-> the \textbf{\acrfull{acr::lod}}; + \item<3-> the \textbf{\emph{finesse}}: the semantic evaluation level. \end{itemize} ~\\ \uncover<4->{ @@ -223,7 +223,7 @@ \end{figure} \end{frame} - \begin{frame}{The evalutation pipeline sketch} + \begin{frame}{The evaluation pipeline sketch} \begin{figure} \includegraphics[width=\textwidth]{graphical_abstract} \end{figure} @@ -252,26 +252,51 @@ \end{frame} \section{Experiments} - \begin{frame}{Used data} + \begin{frame}{Results using all features} \begin{table} \begin{center} + \scriptsize \begin{tabular}{c c c c} \toprule - & Elancourt & Nantes & Paris 13 \\ + & Elancourt & Nantes & Paris 13 \\ \midrule \# samples & 2009 & 748 & 478 \\ \bottomrule \end{tabular} + \caption{\scriptsize Dataset statistics} + \uncover<2->{ + \begin{tabular}{|x{1cm} | x{1.3cm} x{1.3cm} | x{1.2cm} x{1.2cm} | x{1.2cm} x{1.2cm}|} + \hline + & \multicolumn{2}{x{3cm}|}{\textbf{Elancourt (10-cross val.)}} & \multicolumn{2}{x{2.4cm}|}{\textbf{Elancourt $\rightarrow$ Nantes}} & \multicolumn{2}{x{2.5cm}|}{\textbf{Elancourt $\rightarrow$ Paris 13}}\\ + \cline{2-7} + &\textbf{Recall} & \textbf{Prec.} & \textbf{Recall} & \textbf{Prec.} & \textbf{Recall} & \textbf{Prec.} \\ + \hline + \textit{BOS} & 90.83 & 76.14 & \textcolor{IGNDarkGreen}{93.12} & \textcolor{red}{42.61} & \textcolor{IGNDarkGreen}{96.53} & \textcolor{red}{43.82} \\ + \hline + \textit{BUS} & 39.32 & 71.81 & \textcolor{red}{8.82} & 66.67 & \textcolor{red}{0} & \textcolor{red}{---} \\ + \hline + \textit{BImB} & 16.75 & 68.0 & \textcolor{red}{2.02} & 33.33 & \textcolor{red}{0} & \textcolor{red}{---} \\ + \hline + \textit{BInT} & 11.11 & 91.67 & 0.88 & 100 & 3.95 & 50.0 \\ + \hline + \hline + \textit{FOS} & 98.91 & 98.84 & \textcolor{IGNDarkGreen}{98.33} & \textcolor{IGNDarkGreen}{97.92} & \textcolor{IGNDarkGreen}{97.19} & \textcolor{IGNDarkGreen}{97.58} \\ + \hline + \textit{FUS} & 1.27 & 66.67 & \textcolor{IGNDarkGreen}{13.81} & \textcolor{IGNDarkGreen}{63.04} & 8.36 & 95.83 \\ + \hline + \textit{FImB} & 7.42 & 100 & \textcolor{IGNDarkGreen}{46.34} & 65.52 & 11.80 & 60.71 \\ + \hline + \textit{FImT} & 3.33 & 100 & 9.09 & 100 & 0 & 0 \\ + \hline + \textit{FIG} & 79.02 & 71.82 & \textcolor{IGNDarkGreen}{94.17} & 70.70 & \textcolor{IGNDarkGreen}{86.16} & \textcolor{IGNDarkGreen}{88.47} \\ + \hline + \end{tabular} + \caption{\scriptsize Test results using \textbf{Random Forest} ($max\ depth = 4$ \& $\# trees = 1000$) trained on Elancourt.} + } \end{center} \end{table} \end{frame} - \begin{frame}{Results} - - \end{frame} - \begin{frame}{Discussion} - - \end{frame} - \begin{frame}{Some failure cases} + \begin{frame}{Some failure cases on Elancourt} \begin{figure} \begin{center} \tiny @@ -282,9 +307,9 @@ \textbf{Errors} & \textbf{G.T.} & \textbf{Pred.} & \textbf{Errors} & \textbf{G.T.} & \textbf{Pred.} & \textbf{Errors} & \textbf{G.T.} & \textbf{Pred.} & \textbf{Errors} & \textbf{G.T.} & \textbf{Pred.}\\ \hline \textit{BOS} & \xmark & \cmark & \textit{BUS} & \xmark & \cmark & \textit{BOS} & \cmark & \cmark & \textit{BOS} & \cmark & \xmark \\ - Valid & \cmark & \xmark & \textit{FImS} & \cmark & \xmark & \textit{FUS} & \cmark & \xmark & \textit{FOS} & \cmark & \xmark \\ + Valid & \cmark & \xmark & \textit{FIG} & \cmark & \xmark & \textit{FUS} & \cmark & \xmark & \textit{FOS} & \cmark & \xmark \\ & & & \textit{FOS} & \cmark & \xmark & & & & \textit{BUS} & \cmark & \xmark \\ - & & & & & & & & & \textit{BInF} & \cmark & \cmark \\ + & & & & & & & & & \textit{BImB} & \cmark & \cmark \\ \hline \end{tabular} \end{center} @@ -292,13 +317,19 @@ \end{frame} \section{Conclusion} - \begin{frame}{Conclusion \& Perspectives} - \begin{itemize}[label=$\blacktriangleright$, font=\color{IGNGreen}, itemsep=2em] - \item \textbf{Flexible, robust and hierarchical} taxonomy agnostic to inputs; + \begin{frame}{\textcolor{yellow}{Conclusion} \& \textcolor{purple}{Perspectives}} + \begin{itemize}[label=$\blacktriangleright$, font=\color{yellow}, itemsep=2em] + \item \textbf{Flexible, robust and hierarchical} taxonomy; \item \textbf{Fast, lightweight and modular} pipeline for model evaluation; \item Baseline for geometric, image-based and height-based features; - \item More annotated data $\longrightarrow$ \textbf{Simulate errors} from the taxonomy based on reference data; - \item Need for features leveraging data structure $\longrightarrow$ Graph kernels, deep learning. \end{itemize} + \uncover<2->{ + ~\\ + Future work: + \begin{itemize}[label=$\blacktriangleright$, font=\color{purple}, itemsep=2em] + \item Dataset augmentation $\longrightarrow$ \textbf{Simulate errors} from the taxonomy based on reference data; + \item Extend to richer features $\longrightarrow$ Graph kernels, deep learning. + \end{itemize} + } \end{frame} \end{document} \ No newline at end of file