Getting black image! about imgaug HOT 7 OPEN

caffe's load_image seems to return floats in the range 0-1.0, while the library usually expects uint8 in the range 0-255. So you would have to do something like (img*255).astype(np.uint8) after loading the images.
Considering your image size of 32x32, the number of augmenters is also rather large. You should probably decrease each augmenters probability by changing st = lambda aug: iaa.Sometimes(0.3, aug) to st = lambda aug: iaa.Sometimes(0.1, aug).
Further, you should use something like seq.show_grid(images[0], rows=8, cols=8) to show your augmenters results instead of plt.imshow(images_aug[0]), as the latter one will result in the same augmentation at every script run.

from imgaug.

I meant a command line program see, not a pypi package with that name, i.e. you must be able to type see in the command line and not get an error (that's how I understood it from the help threads - the program seems to be available on Ubuntu by default).

from imgaug.

OK, I ran some more tests individually on each of those options, it seems onlyiaa.Fliplr(0.5)and iaa.Flipud(0.5) work. the rest result in a black image which I posted above!
What am I missing here?!
(By the way I'm on windows10.)

from imgaug.

@aleju : Thank you very much. The black image issue is gone.
and speaking of seg.show_grid(), it fails for me with this error RuntimeError: Could not execute image viewer.
I could'nt find a way to solve it, I have no idea whats causing this! so thats why I used the plt.imshow instead

from imgaug.

The threads that I can find sound like the command see must be available for misc.imshow to work (or an image viewer must be set in the environment variable SCIPY_PIL_IMAGE_VIEWER). I guess that command isn't available in Windows 10. Maybe I'll switch to OpenCV's imshow equivalent.

from imgaug.

I installed see through pip yet I still get the same error.

from imgaug.

how to install the see program over mac os

from imgaug.