Trouble using CSBDeep with Fiji

Hi all, I am trying to use CSBDeep > N2V in Fiji. But keep getting an error: ‘A RandomAccessibleInterval is required but none exist’. Any idea what might be the problem? (this applies for DenoiseSeg). And when I tried CSBDeep > Demo, there is an error ‘A dataset is required but none exist’. I tried this on both Mac and Windows and the same problem persist. I have updated my Fiji to the latest version. Any help is much appreciated!

1 Like

Hi @ttan ,

the error suggests you don’t have any image open when clicking the menu.

Can you try opening an image first?


1 Like

@haesleinhuepf It works, thanks!

1 Like

Hitting another snag. I can use N2V > N2V train now and save the model. But when I try to do prediction, I get error messages from Console ‘[ERROR] Module threw error
java.lang.NoClassDefFoundError: net/imglib2/cache/img/optional/CacheOptions$CacheType’. Any idea what to try? Thanks!

Tried running this on a Windows. I am able to train the model fine, but prediction steps continue to be a problem. It seems to be throwing a different error though:

Hey hey,

Is your Fiji up to date (Help / Update…)? This might be linked to an error fixed in the most recent imagej-modelzoo.jar, which you might not have downloaded yet through the updater.


Dear all, I thought I would use this thread to mention a recurrent problem I’ve been having and would love to discuss with you and have hints as to how to correct this.

When using TF 1.15 GPU on Fiji to train either DenoiSeg or N2V models, my Fiji just quits everytime.

It does not happen while I am on the PC, but seems to get triggered when working or connecting remotely.

While the PC is on and I am working with it, everything works fine but if I disconnect a remote session and reconnect, then Fiji is no longer running and I have to restart it. Any ideas om how I could go about debugging this?

Note that running the training on Python within the same PC causes no problems.

Some Specs: RTX 2080 GPU with 8GB or RAM on Windows 10 build 20H2