You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
). Our imager automatically uses .TIF, which means we have to rename everything to .tif to get images to work with cellpose. It's not the end of the world, but seems like an easy-ish fix on the cellpose side.
More to the point, I spent a long time trying to figure out errors during training via the command line ('ERROR: no images in --dir folder'). The directory wasn't empty, but it was full of images named .TIF. I had to dig into the code to find what was going on. I think there are a few options:
Make clear in the documentation that image extensions are case-sensitive.
Include a more informative error in get_image_files() when no images are found.
Add more potential extensions in get_image_files()
Maybe switch to regex instead of glob? One problem with glob is it's going to behave differently across Windows/Linux. See here.
Add a new argument that allows the user to provide the extension.
Using images with .TIF extensions hasn't caused me any problems elsewhere (i.e., cv2, scikit-image, etc.) so I doubt allowing them to be read would cause problems downstream, but I can't say for sure.
Thanks for the great tool! I'm excited to integrate it into our pipelines. Cheers.
The text was updated successfully, but these errors were encountered:
The IO script is case-sensitive for image extensions (see:
cellpose/cellpose/io.py
Line 138 in 91dd7ab
More to the point, I spent a long time trying to figure out errors during training via the command line (
'ERROR: no images in --dir folder'
). The directory wasn't empty, but it was full of images named .TIF. I had to dig into the code to find what was going on. I think there are a few options:get_image_files()
when no images are found.get_image_files()
Using images with .TIF extensions hasn't caused me any problems elsewhere (i.e., cv2, scikit-image, etc.) so I doubt allowing them to be read would cause problems downstream, but I can't say for sure.
Thanks for the great tool! I'm excited to integrate it into our pipelines. Cheers.
The text was updated successfully, but these errors were encountered: