diff --git a/trunk/RTCSA_SS/03Design.tex b/trunk/RTCSA_SS/03Design.tex
index 15d1660..1660667 100644
--- a/trunk/RTCSA_SS/03Design.tex
+++ b/trunk/RTCSA_SS/03Design.tex
@@ -36,13 +36,13 @@ to two categories. One is lay on back and the other is lay on side. Since the in
 data is very small, we use a neural network consist one 2D convolution layer, one
 2D max pooling, one flatten and one densely-connected layer. The possibility of
 output has a very large various just after turning over because the model cannot
-distinguish the residual heat on bed and the person as Figure~\ref{fig:pose}(a) shown. This
+distinguish the residual heat on bed and the person as Figure~\ref{fig:residual_heat} shown. This
 situation will slowly disappear after one or two minutes.
 
 To determination the pose, first we use a median filter with a window size of five
 to filter out the noise. Than, find the curve hull line of the upper bound and
 lower bound of the data. Finally calculate the middle line of upper bound and lower bound.
-Figure~\ref{fig:pose}(b) and (c) shows the data and these lines.
+Figure~\ref{fig:trend} shows the data and these lines.
 
 We divide every data into 10 second time windows. If the middle line of the time window
 is at the top one fifth, it is a lay on back. If it is at the bottom one fifth,
@@ -52,15 +52,14 @@ there are three continuously same output.
 
 \begin{figure}[ht]
   \centering
-  \subfloat[Residual heat on bed]{
-    \includegraphics[width=0.3\columnwidth]{figures/Lepton_residual_heat.bmp}
-  }
-  \subfloat[Enhanced Images after Background Calibration]{
-    \includegraphics[width=0.3\columnwidth]{figures/MinMax.pdf}
-  }
-  \subfloat[Enhanced Images after Background Calibration]{
-    \includegraphics[width=0.3\columnwidth]{figures/Mid.pdf}
-  }
-  \caption{Background subtraction.}
-  \label{fig:pose}
+  \minipage{0.3\columnwidth}
+    \includegraphics[width=\linewidth]{figures/Lepton_residual_heat.png}
+    \caption{Residual heat on bed.}
+    \label{fig:residual_heat}
+  \endminipage
+  \minipage{0.65\columnwidth}
+    \includegraphics[width=\linewidth]{figures/MinMax.pdf}
+    \caption{Trend of pose.}
+    \label{fig:trend}
+  \endminipage
 \end{figure}
diff --git a/trunk/RTCSA_SS/04Evaluation.tex b/trunk/RTCSA_SS/04Evaluation.tex
index 8a7b4ef..d0d186a 100644
--- a/trunk/RTCSA_SS/04Evaluation.tex
+++ b/trunk/RTCSA_SS/04Evaluation.tex
@@ -1,4 +1,30 @@
 \section{Performance Evaluation}
 \label{sec:eval}
 
-Evaluation.
\ No newline at end of file
+This section presents the evaluation results for the proposed method, and
+how we collect the dataset.
+
+For training SRCNN model, we let a person lay on bed and randomly change pose
+or move his arms and legs. Collect the about 600 images from Grideye sensors and
+Lepton 3, and align them at the same timestamps. Figure~\ref{fig:resolution_compare} shows the result of SRCNN model.
+
+\begin{figure}[ht]
+	\centering
+	\subfloat[Grideye Image]{
+		\includegraphics[width=0.32\columnwidth]{figures/LR.png}
+	}
+	\subfloat[SR Image]{
+		\includegraphics[width=0.32\columnwidth]{figures/SR.png}
+	}
+	\subfloat[Downscaled Lepton Image]{
+		\includegraphics[width=0.32\columnwidth]{figures/HR.png}
+	}
+	\caption{Result of SRCNN}
+	\label{fig:resolution_compare}
+\end{figure}
+
+For training the pose recognization model, we collect 200 images of lay on back and 400 images
+of lay on right or left side. The result shows that the accuracy of single frame detection
+can be improved about 5\% by SRCNN.
+
+The accuracy of pose detection is about 67\% and turning over datection is 56\%.
diff --git a/trunk/RTCSA_SS/05Conclusion.tex b/trunk/RTCSA_SS/05Conclusion.tex
index 4f5b1aa..2b81dc3 100644
--- a/trunk/RTCSA_SS/05Conclusion.tex
+++ b/trunk/RTCSA_SS/05Conclusion.tex
@@ -1,3 +1,9 @@
 \section{Conclusion\label{sec:conclusion}}
 
-Conclusion
\ No newline at end of file
+In this paper, we use the SRCNN to improve the resolution of thermal sensor, and
+detect the pose of each frame.
+The result shows that Super-resolution can slightly improve the accuracy of pose
+detection. We develop a method to detect the turning over. It has about 67\% accuracy
+even when the accuracy of pose detection is only ???\%.
+
+
diff --git a/trunk/RTCSA_SS/Main.tex b/trunk/RTCSA_SS/Main.tex
index d602cf6..48c8837 100644
--- a/trunk/RTCSA_SS/Main.tex
+++ b/trunk/RTCSA_SS/Main.tex
@@ -13,7 +13,6 @@
 
 %\usepackage{ntu_techrpt_cover}
 %\usepackage{lipsum}
-\usepackage{graphicx}
 \usepackage{times}
 %\usepackage{psfrag}
 %\usepackage[tight]{subfigure}
@@ -22,7 +21,8 @@
 %\usepackage{epsfig}
 \usepackage{longtable}
 %\usepackage{cases}
-%\usepackage{subfig}
+\usepackage{subfig}
+\usepackage{graphicx}
 \usepackage{balance}
 \usepackage{xcolor}
 %\usepackage{algorithm}
diff --git a/trunk/RTCSA_SS/figures/HR.png b/trunk/RTCSA_SS/figures/HR.png
new file mode 100644
index 0000000..5c02b5d
Binary files /dev/null and b/trunk/RTCSA_SS/figures/HR.png differ
diff --git a/trunk/RTCSA_SS/figures/LR.png b/trunk/RTCSA_SS/figures/LR.png
new file mode 100644
index 0000000..f3abee4
Binary files /dev/null and b/trunk/RTCSA_SS/figures/LR.png differ
diff --git a/trunk/RTCSA_SS/figures/Lepton_residual_heat.bmp b/trunk/RTCSA_SS/figures/Lepton_residual_heat.bmp
deleted file mode 100644
index 9eca0c1..0000000
Binary files a/trunk/RTCSA_SS/figures/Lepton_residual_heat.bmp and /dev/null differ
diff --git a/trunk/RTCSA_SS/figures/Lepton_residual_heat.png b/trunk/RTCSA_SS/figures/Lepton_residual_heat.png
new file mode 100644
index 0000000..4941029
Binary files /dev/null and b/trunk/RTCSA_SS/figures/Lepton_residual_heat.png differ
diff --git a/trunk/RTCSA_SS/figures/SR.png b/trunk/RTCSA_SS/figures/SR.png
new file mode 100644
index 0000000..f86d916
Binary files /dev/null and b/trunk/RTCSA_SS/figures/SR.png differ
diff --git a/trunk/RTCSA_SS/figures/coverart.eps b/trunk/RTCSA_SS/figures/coverart.eps
new file mode 120000
index 0000000..ca78d61
--- /dev/null
+++ b/trunk/RTCSA_SS/figures/coverart.eps
@@ -0,0 +1 @@
+/Users/cshih/notes/tex_config/figures/coverart.eps
\ No newline at end of file