search

miliadis / DeepVideoCS

PyTorch deep learning framework for video compressive sensing.
BSD 2-Clause "Simplified" License
74 stars 22 forks source link

An error aroused by API change in 'imageio' #5

Closed Ericgeese closed 3 years ago

Ericgeese commented 3 years ago

I just want to point out a small error in line 50 of datasets/videocs.py The error information shows that "'float' object cannot be interpreted as an integer". According to https://stackoverflow.com/questions/54778001/how-to-to-tackle-overflowerror-cannot-convert-float-infinity-to-integer, it is caused by an update in 'imageio' in 2019. The solution is to replace reader._meta['nframes'] with reader.count_frames()

balablb commented 3 years ago

thanks a lot , I have the same problem

balablb commented 3 years ago

I just want to point out a small error in line 50 of datasets/videocs.py The error information shows that "'float' object cannot be interpreted as an integer". According to https://stackoverflow.com/questions/54778001/how-to-to-tackle-overflowerror-cannot-convert-float-infinity-to-integer, it is caused by an update in 'imageio' in 2019. The solution is to replace reader._meta['nframes'] with reader.count_frames()

Have you ever had this problem? Test: [1/1] Time 97.376 (97.376) PSNR -26.118 (-26.118) Lossy conversion from float64 to uint8. Range [-0.04140618070960045, 255.0]. Convert image to uint8 prior to saving to suppress this warning.

miliadis commented 3 years ago

Thanks, I modernized all the code so it will now work seamlessly with Ubuntu 18.04, Python 3.9.1, PyTorch 1.7.1 and CUDA 11.0. I have also fixed the "Convert image to uint8 prior to saving to suppress this warning." message.