Python / Opencv Float在转换时转换为不正确的Int

Question

I am using OpenCV + Python to add an overlay image on a few videos automatically. For the most part my code works fine, except with collecting some heuristics from a given input video. I collect the number of frames and the FPS of the video using the OpenCV VideoCapture.Get() function and store then to variables respectively. The resulting numeric are float values which I then cast to int , but here lies the problem: the values drop by one!

The full code snippet here:

import cv2 as cv

inputnm = "testVideo.mp4"

# Load the video into memory
camera = cv.VideoCapture(inputnm)

# Get important heuristics
fps = camera.get(cv.CAP_PROP_FPS)
maxFrames = camera.get(cv.CAP_PROP_FRAME_COUNT)

print type(fps), type(maxFrames)
print fps,maxFrames
print int(fps), int(maxFrames)
quit()

This produces the output:

<type 'float'> <type 'float'>
60.0 401.0
59 400

Can anyone explain this to me why? I have tried converting it to numpy.float128 and then to numpy.unit8 but it still drops the values down by one. Strange.

Answer 1

You have just seen the wonders of the floating point.

Python shows numbers as a close precision to their conversation from floating point, which is usually precise enough for visualization and in many cases also for computation, but in cases like yours the precision is simply not enough.

Use round(maxFrames) .

Answer 2

Essentially what is happening is your floating point variable is displaying as 60, but in reality it is probably 59 and some change. It will not display as 59 because python is smart enough to round up for our sake, (59.99999.. = 60). When you cast to int, you're truncating the decimals off of 59 which seems to drop the value by 1 (59.9999 => 59).

Edit: like the other user suggested, round(maxFrames) will round the value to the nearest whole number. Then you can cast to int and get the right result.