This is a test program to see how many FPS the camera is capturing at

with OpenCV. It will print the results once a second. Hit Control-C to
cancel.

With my particular camera, every frame I read makes the next frame
even slower. This doesn't make any sense. It's definitely a bug in
OpenCV as it gets fast again if I release and then reopen the device.
Could it be a bug in the JPEG decompression routines?

opencvtest.py

@@ -0,0 +1,75 @@
+#!/usr/bin/python
+
+# What the heck. My Logitech QuickCam gets slower on every read() when
+# I use OpenCV 2.4.9 but works fine with other V4L2 apps. If I release
+# and reopen the camera, it becomes fast again.
+
+# This program just prints out FPS every second. This is what the
+# output looks like on my Raspberry Pi 3B and Logitech QuickCam:
+#
+#####################################################################
+# ./opencvtest.py                                                   #
+# Camera detected as 640 x 480                                      #
+# First frame took 3.501s to grab                                   #
+# After 0016 frames, 0001 seconds: Camera is capturing at 14.26 fps #
+# After 0031 frames, 0002 seconds: Camera is capturing at 14.20 fps #
+# After 0044 frames, 0003 seconds: Camera is capturing at 12.69 fps #
+# After 0056 frames, 0004 seconds: Camera is capturing at 11.37 fps #
+# After 0067 frames, 0005 seconds: Camera is capturing at 10.11 fps #
+# After 0077 frames, 0006 seconds: Camera is capturing at 9.47 fps  #
+# After 0086 frames, 0007 seconds: Camera is capturing at 8.79 fps  #
+# After 0095 frames, 0008 seconds: Camera is capturing at 8.27 fps  #
+# After 0103 frames, 0009 seconds: Camera is capturing at 7.81 fps  #
+# ...                                                               #
+# After 0199 frames, 0025 seconds: Camera is capturing at 4.88 fps  #
+# After 0204 frames, 0026 seconds: Camera is capturing at 4.73 fps  #
+# ...                                                               #
+# After 0298 frames, 0050 seconds: Camera is capturing at 3.47 fps  #
+# After 0302 frames, 0051 seconds: Camera is capturing at 3.47 fps  #
+# ...                                                               #
+# After 0399 frames, 0083 seconds: Camera is capturing at 2.75 fps  #
+# After 0402 frames, 0084 seconds: Camera is capturing at 2.67 fps  #
+# ...                                                               #
+# After 0498 frames, 0153 seconds: Camera is capturing at 0.50 fps  #
+# After 0500 frames, 0156 seconds: Camera is capturing at 0.99 fps  #
+# After 0502 frames, 0157 seconds: Camera is capturing at 1.33 fps  #
+# After 0504 frames, 0159 seconds: Camera is capturing at 0.99 fps  #
+# After 0506 frames, 0166 seconds: Camera is capturing at 0.30 fps  #
+#####################################################################
+
+import time
+import cv2
+
+# Open the first available camera
+cap = cv2.VideoCapture(-1)
+if not cap.isOpened:
+    print "Warning: Failed to open camera using OpenCV"
+    exit
+
+# Print the capabilities of the connected camera
+w=cap.get(cv2.cv.CV_CAP_PROP_FRAME_WIDTH)
+h=cap.get(cv2.cv.CV_CAP_PROP_FRAME_HEIGHT)
+if w and h:
+    print("Camera detected as %d x %d" % (w, h))
+
+# The first read always takes a long time. No idea why.
+start = time.time()
+cap.grab()
+end = time.time()
+print("First frame took %0.3fs to grab" % (end - start) )
+
+overall_start=time.time()
+total_frames=1
+while True:
+        # Measure actual FPS of camera
+        frames=0
+        start = time.time()
+        while (time.time() - 1 < start):
+            frames = frames + 1
+            cap.grab()          # <== This is where the bottleneck is
+        end = time.time()
+        fps = frames/(end-start)
+        total_frames=total_frames+frames
+        print("After %04d frames, %04d seconds: Camera is capturing at %.2f fps" % (total_frames, int(time.time()-overall_start), fps))
+
+