\begin{tabular}{c | c | c | c | c}
{\bf Method} & {\bf Moderate} & {\bf Easy} & {\bf Hard} & {\bf Runtime}\\ \hline
SubCNN \cite{xiang2017subcategory} & 63.41 \% & 71.39 \% & 56.34 \% & 2 s / GPU \\
Deep3DBox \cite{MousavianCVPR2017} & 59.37 \% & 68.58 \% & 51.97 \% & 1.5 s / GPU \\
3DOP \cite{Chen2015NIPS} & 58.59 \% & 71.95 \% & 52.35 \% & 3s / GPU \\
DeepStereoOP \cite{Pham2017SPIC} & 55.62 \% & 67.49 \% & 48.85 \% & 3.4 s / GPU \\
Mono3D \cite{Chen2016CVPR} & 53.11 \% & 65.74 \% & 48.87 \% & 4.2 s / GPU \\
FRCNN+Or \cite{GuindelICVES} & 51.47 \% & 64.90 \% & 46.48 \% & 0.1 s / GPU \\
maxFtr+ROI \cite{Tian2017VISAPP} & 38.29 \% & 42.96 \% & 34.28 \% & 0.25 s / 4 cores \\
DPM-VOC+VP \cite{Pepik2015PAMI} & 23.22 \% & 31.24 \% & 21.62 \% & 8 s / 1 core \\
LSVM-MDPM-sv \cite{Felzenszwalb2010PAMI} & 23.14 \% & 28.89 \% & 22.28 \% & 10 s / 4 cores \\
DPM-C8B1 \cite{Yebes2015SENSORS} & 19.25 \% & 27.16 \% & 17.95 \% & 15 s / 4 cores
\end{tabular}