removed savgol filter from handpose

HandPose/handPoseVideo.py

@@ -1,8 +1,7 @@
 import cv2
 import time
 import numpy as np
-import scipy
-from scipy import signal
+
 
 protoFile = "hand/pose_deploy.prototxt"
 weightsFile = "hand/pose_iter_102000.caffemodel"
@@ -11,6 +10,7 @@
 
 threshold = 0.2
 
+
 input_source = "asl.mp4"
 cap = cv2.VideoCapture(input_source)
 hasFrame, frame = cap.read()
@@ -27,22 +27,13 @@
 
 net = cv2.dnn.readNetFromCaffe(protoFile, weightsFile)
 k = 0
-
-smoothed_data, data, previous_points, frame_number = [], [], [], 0
-for i in range(nPoints): previous_points.append((0,0))
-# Smoothing parameters - choose a window length of about 1/2 to 1/4 the fps (so 60 fps --> window length of 33)
-window_length, exponent_value = 33,2 # Window length must be odd
-
-### This is going to take some time before the party get's started! The downside of smoothing is that
-### data from the past and the future is required to smooth data in the present. This means that all the frames
-### in the video must be processed before smoothing the data and displaying the result. This method is therefore
-### not suitable for realtime results.
 while 1:
     k+=1
     t = time.time()
     hasFrame, frame = cap.read()
     frameCopy = np.copy(frame)
     if not hasFrame:
+        cv2.waitKey()
         break
 
     inpBlob = cv2.dnn.blobFromImage(frame, 1.0 / 255, (inWidth, inHeight),
@@ -72,64 +63,30 @@
             # Add the point to the list if the probability is greater than the threshold
             points.append((int(point[0]), int(point[1])))
         else :
-            # Add the last known point (ex: if thumb is not detected, use thumb position from previous frame)
-            points.append(previous_points[i])
-
-    # Save the data from the model - data is a list of lists. Each element is a list containing the 22 coordinates for the hand.
-    data.append(points)
-
-    previous_points = points
-
-    print("total = {}".format(time.time() - t))
-
-# Re-capture the source, so that the video starts at the beginning again
-cap = cv2.VideoCapture(input_source)
+            points.append(None)
 
-# Smooth it out
-# Split the data so that just the x values for the first point are made into a list
-smoothed_data_in_series = []
-for point_index in range(nPoints): # Iterate through each point (wrist, thumb, etc...)
-    data_point_series_x = []
-    data_point_series_y = []
-    for values in data: # Iterate through the series of data (each frame of video)
-        data_point_series_x.append(values[point_index][0])
-        data_point_series_y.append(values[point_index][1])
-    # Smooth the x and y values
-    smoothed_data_point_series_x = signal.savgol_filter(data_point_series_x, window_length, exponent_value)
-    smoothed_data_point_series_y = signal.savgol_filter(data_point_series_y, window_length, exponent_value)
-    smoothed_data_in_series.append(smoothed_data_point_series_x)
-    smoothed_data_in_series.append(smoothed_data_point_series_y)
-
-# Now the data is sepearted into 44 lists (two lists for each of the 22 locations of the hand, time to zip()
-for current_frame_number in range(len(smoothed_data_in_series[0])):
-    frame_values = []
-    for point_index in range(nPoints):
-        x = smoothed_data_in_series[point_index*2][current_frame_number]
-        y = smoothed_data_in_series[point_index*2+1][current_frame_number]
-        frame_values.append((x,y))
-    smoothed_data.append(frame_values)
-
-# Iterate through each frame of data
-for values in smoothed_data:
-
-    hasFrame, img = cap.read()
-
-    # When data is smoothed, floats are introduced, these must be eliminated becase cv2.circle requries integers
-    values = np.array(values, int)  
-
     # Draw Skeleton
     for pair in POSE_PAIRS:
         partA = pair[0]
         partB = pair[1]
-        cv2.line(img, (values[partA][0], values[partA][1]),(values[partB][0], values[partB][1]), (32, 255, 0), 2, lineType=cv2.LINE_AA)
-        cv2.circle(img, (values[partA][0], values[partA][1]), 2, (0, 0, 255), thickness=-1, lineType=cv2.FILLED)
-        cv2.circle(img, (values[partB][0], values[partB][1]), 2, (0, 0, 255), thickness=-1, lineType=cv2.FILLED)
-
-    cv2.imshow('Output-Skeleton', img)
-    cv2.waitKey(0)
-    vid_writer.write(img)
-
-cv2.destroyAllWindows()
-vid_writer.release()
-cap.release()
 
+        if points[partA] and points[partB]:
+            cv2.line(frame, points[partA], points[partB], (0, 255, 255), 2, lineType=cv2.LINE_AA)
+            cv2.circle(frame, points[partA], 5, (0, 0, 255), thickness=-1, lineType=cv2.FILLED)
+            cv2.circle(frame, points[partB], 5, (0, 0, 255), thickness=-1, lineType=cv2.FILLED)
+
+    print("Time Taken for frame = {}".format(time.time() - t))
+
+    # cv2.putText(frame, "time taken = {:.2f} sec".format(time.time() - t), (50, 50), cv2.FONT_HERSHEY_COMPLEX, .8, (255, 50, 0), 2, lineType=cv2.LINE_AA)
+    # cv2.putText(frame, "Hand Pose using OpenCV", (50, 50), cv2.FONT_HERSHEY_COMPLEX, 1, (255, 50, 0), 2, lineType=cv2.LINE_AA)
+    cv2.imshow('Output-Skeleton', frame)
+    # cv2.imwrite("video_output/{:03d}.jpg".format(k), frame)
+    key = cv2.waitKey(1)
+    if key == 27:
+        break
+
+    print("total = {}".format(time.time() - t))
+
+    vid_writer.write(frame)
+
+vid_writer.release()