Closed eldon922 closed 2 months ago
What do you mean by "slightly different"? Could the difference be caused by different versions of NumPy or OpenCV? Check the build log to find which versions your app is using – you can force pip to run again by making any change to the pip
block in your build.gradle file. Then try using those same versions on Windows.
What do you mean by "slightly different"? Could the difference be caused by different versions of NumPy or OpenCV? Check the build log to find which versions your app is using – you can force pip to run again by making any change to the
pip
block in your build.gradle file. Then try using those same versions on Windows.
The difference lies in the numbers; they are only slightly different. I don't believe this discrepancy is caused by variations in dependencies, as the values differ between the real device and emulator, both of which are using the exact same build.
I believe this is caused by different codecs that used for reading image? Reference. Are there any ways to set what codecs used?
So you've tried 3 devices: Android emulator, Android table, and Windows. Are all 3 of them different, or are 2 of them the same? If the Android emulator is the same as Windows, but the Android tablet is different, then it could be an x86 vs ARM difference.
Can you attach an image file which shows the issue?
Sorry, it turns out the bitmap compress function that i use to copy the image caused this discrepancy issue. Thank you for the response
Chaquopy version
15.0.1
Devices or emulators where the issue happens
Real Android tablet devices
Relevant parts of your code
cv2.imread(join(os.environ["HOME"], imagePath))
Describe your issue
I write the array values to a text file and compare the results between them. However, they are slightly different, which affects our subsequent processes. Is there an alternative function that ensures consistent values regardless of the machine used?