Docker template updated (only the read_image function in the inference.py file)

For those who have problems in creating and/or submitting their docker or those who just want to start ;

I've updated the read_image function in the inference.py file thanks to the Problems in submission forum channel** and @danifranco's contribution.

You can either update it by downloading the updated Docker template here or by visiting the code in the forum topic here.

How do you test the algorithm before making a submission ('Try-out algorithm' menu)?

1/ If I upload one of the angle files we have in the training set, I get "Cancelled, One or more of the inputs failed validation." (Result ID a74d42b6-47a2-4fb5-8f4b-982e717f0700).

2/ If I use one of the pre-uploaded tif files of Grand Challenge, I get:

2025-02-24T14:41:37.367000+00:00 Traceback (most recent call last):
2025-02-24T14:41:37.367000+00:00   File "/opt/app/inference.py", line 130, in <module>
2025-02-24T14:41:37.368000+00:00     raise SystemExit(run())
2025-02-24T14:41:37.368000+00:00                      ^^^^^
2025-02-24T14:41:37.368000+00:00   File "/opt/app/inference.py", line 54, in run
2025-02-24T14:41:37.368000+00:00     save_image(
2025-02-24T14:41:37.368000+00:00   File "/opt/app/inference.py", line 117, in save_image
2025-02-24T14:41:37.368000+00:00     PhysicalSizeX = metadata['PhysicalSizeX']
2025-02-24T14:41:37.368000+00:00                     ~~~~~~~~^^^^^^^^^^^^^^^^^
2025-02-24T14:41:37.368000+00:00 KeyError: 'PhysicalSizeX'

So I imagine the pre-uploaded file by the platform don't have metadata (or not the same as the Fuse My Cells challenge images)?

So I'm a bit confused, have I missed something?

Hello,

This Grand Challenge documentation may help you test your algorithm https://grand-challenge.org/documentation/try-out-your-algorithm/.

It seems that you have tried your algorithm on another type of image that does not correspond to our challenge. I invite you to choose “upload a new image” and take an image from the training database. As you have 5 submissions for the test phase, you can also try with the test images.

Did it help?

Hi, as mentioned previously, if I upload one of the angle files we have in our Fuse My Cells training set, I get "Cancelled, One or more of the inputs failed validation."

More details:

• Uploading: https://i.imgur.com/h70LQFq.png

• Error: https://i.imgur.com/1R5AvJi.png

Do you have logs?
It's a naive question, but have you checked that the entries/inputs are well being read in your docker?
Because it looks like the image is being imported thanks to your images, but as no entries/inputs have been found, it can't be working.

I get "No logs found on stdout" and "No logs found on stderr". I don't think I've changed anything docker related compared to the template you provided, except I'm using Python 3.12. But if it was code-related I would have logs right?

Especially since it works with the pre-uploaded images of Grand Challenge (they just don't have the metadata we have here for FuseMyCells).

Sorry I'm not very good with Docker so I don't know how to provide more information.

EDIT: it might be a problem with my code, weird that I didn't have a stderr. Currently looking into it. Sorry.

Hi JasonMendoza2008,

I contacted Grand Challenge for further clarification and tested the try-out algorithm with our test algorithm. Thanks to their help and efforts, it turns out that the selected image (image_189_nucleus_angle.tif) does not pass Grand Challenge’s basic validation.

As a result, you can upload your image, but it cannot be read or imported.

When I tried, I received the following message:

Image validation for interface Fluorescence Lightsheet 3D Microscopy failed with error: 1 file could not be imported.

It seems that this issue stems from our writer. We used imwrite from tifffile, and it appears that when we set the parameter imagej=True(which we did for the training database), Grand Challenge is unable to import and validate the image.

I've rewritten the image with imagej=False, but the image you chose is quite heavy (1.2Go) so I've rewritten another one lighter, (i.e. 2 images in total)

you can download them here¶

I've tested and it works :

I hope this could help you, do not hesitate if you have other issues.

Sincerely,
Dorian

Hi, that's great, thank you so much for the investigation. I'll use both the images you provided for validation from now on.

Current link "https://seafile.lirmm.fr/d/a952ce4f99484be2ad72/" seems dead :

"/Fusemycells/docker_template/" n'existe pas.

Thank you for the post!

I don't know what happened, here is the new link (updated also on the docker page of the website)

https://seafile.lirmm.fr/d/233a5a399c8544dfb41a/¶