From 93b809d65085460dc3a26b60d07f1f387efa146b Mon Sep 17 00:00:00 2001
From: vikas <vikas.7d@gmail.com>
Date: Thu, 27 Sep 2018 22:46:58 +0530
Subject: [PATCH] added smart resizing for input blob

---
 HandPose/handPoseImage.cpp |  8 ++++++--
 HandPose/handPoseImage.py  | 12 +++++-------
 HandPose/handPoseVideo.cpp |  7 +++++--
 HandPose/handPoseVideo.py  | 13 ++++++++-----
 4 files changed, 24 insertions(+), 16 deletions(-)

diff --git a/HandPose/handPoseImage.cpp b/HandPose/handPoseImage.cpp
index 7b05bc7e0..680f6f421 100644
--- a/HandPose/handPoseImage.cpp
+++ b/HandPose/handPoseImage.cpp
@@ -34,8 +34,6 @@ int main(int argc, char **argv)
       imageFile = argv[1];
     }
 
-    int inWidth = 368;
-    int inHeight = 368;
     float thresh = 0.01;
 
     Mat frame = imread(imageFile);
@@ -43,6 +41,12 @@ int main(int argc, char **argv)
     int frameWidth = frame.cols;
     int frameHeight = frame.rows;
 
+    float aspect_ratio = frameWidth/(float)frameHeight;
+    int inHeight = 368;
+    int inWidth = (int(aspect_ratio*inHeight) * 8) / 8;
+
+    cout << "inWidth = " << inWidth << " ; inHeight = " << inHeight << endl;
+
     double t = (double) cv::getTickCount();
     Net net = readNetFromCaffe(protoFile, weightsFile);
 
diff --git a/HandPose/handPoseImage.py b/HandPose/handPoseImage.py
index 3de3598bf..2d0620ac6 100644
--- a/HandPose/handPoseImage.py
+++ b/HandPose/handPoseImage.py
@@ -6,20 +6,21 @@
 weightsFile = "hand/pose_iter_102000.caffemodel"
 nPoints = 22
 POSE_PAIRS = [ [0,1],[1,2],[2,3],[3,4],[0,5],[5,6],[6,7],[7,8],[0,9],[9,10],[10,11],[11,12],[0,13],[13,14],[14,15],[15,16],[0,17],[17,18],[18,19],[19,20] ]
-
+net = cv2.dnn.readNetFromCaffe(protoFile, weightsFile)
 
 frame = cv2.imread("hand.jpg")
 frameCopy = np.copy(frame)
 frameWidth = frame.shape[1]
 frameHeight = frame.shape[0]
-threshold = 0.1
+aspect_ratio = frameWidth/frameHeight
 
-net = cv2.dnn.readNetFromCaffe(protoFile, weightsFile)
+threshold = 0.1
 
 t = time.time()
 # input image dimensions for the network
-inWidth = 368
 inHeight = 368
+inWidth = int(((aspect_ratio*inHeight)*8)//8)
+
 inpBlob = cv2.dnn.blobFromImage(frame, 1.0 / 255, (inWidth, inHeight),
                           (0, 0, 0), swapRB=False, crop=False)
 
@@ -28,9 +29,6 @@
 output = net.forward()
 print("time taken by network : {:.3f}".format(time.time() - t))
 
-H = output.shape[2]
-W = output.shape[3]
-
 # Empty list to store the detected keypoints
 points = []
 
diff --git a/HandPose/handPoseVideo.cpp b/HandPose/handPoseVideo.cpp
index aee5a435e..0cd3a0c5a 100644
--- a/HandPose/handPoseVideo.cpp
+++ b/HandPose/handPoseVideo.cpp
@@ -23,8 +23,6 @@ int nPoints = 22;
 
 int main(int argc, char **argv)
 {
-    int inWidth = 368;
-    int inHeight = 368;
     float thresh = 0.01;
 
     cv::VideoCapture cap(0);
@@ -38,6 +36,11 @@ int main(int argc, char **argv)
     Mat frame, frameCopy;
     int frameWidth = cap.get(CAP_PROP_FRAME_WIDTH);
     int frameHeight = cap.get(CAP_PROP_FRAME_HEIGHT);
+    float aspect_ratio = frameWidth/(float)frameHeight;
+    int inHeight = 368;
+    int inWidth = (int(aspect_ratio*inHeight) * 8) / 8;
+
+    cout << "inWidth = " << inWidth << " ; inHeight = " << inHeight << endl; 
 
     VideoWriter video("Output-Skeleton.avi",VideoWriter::fourcc('M','J','P','G'), 10, Size(frameWidth,frameHeight));
 
diff --git a/HandPose/handPoseVideo.py b/HandPose/handPoseVideo.py
index 7e6526b38..b34dda948 100644
--- a/HandPose/handPoseVideo.py
+++ b/HandPose/handPoseVideo.py
@@ -8,8 +8,6 @@
 nPoints = 22
 POSE_PAIRS = [ [0,1],[1,2],[2,3],[3,4],[0,5],[5,6],[6,7],[7,8],[0,9],[9,10],[10,11],[11,12],[0,13],[13,14],[14,15],[15,16],[0,17],[17,18],[18,19],[19,20] ]
 
-inWidth = 368
-inHeight = 368
 threshold = 0.1
 
 
@@ -17,6 +15,14 @@
 cap = cv2.VideoCapture(input_source)
 hasFrame, frame = cap.read()
 
+frameWidth = frame.shape[1]
+frameHeight = frame.shape[0]
+
+aspect_ratio = frameWidth/frameHeight
+
+inHeight = 368
+inWidth = int(((aspect_ratio*inHeight)*8)//8)
+
 vid_writer = cv2.VideoWriter('output.avi',cv2.VideoWriter_fourcc('M','J','P','G'), 15, (frame.shape[1],frame.shape[0]))
 
 net = cv2.dnn.readNetFromCaffe(protoFile, weightsFile)
@@ -32,9 +38,6 @@
 
     print("imread = {}".format(time.time() - t))
 
-    frameWidth = frame.shape[1]
-    frameHeight = frame.shape[0]
-
     inpBlob = cv2.dnn.blobFromImage(frame, 1.0 / 255, (inWidth, inHeight),
                               (0, 0, 0), swapRB=False, crop=False)