Kernel crashes when detecting arena

[StevenM711]

Hello!!

I am trying to create a project with the videos and h5 tables that I have, I already ran the tutorial and everything went smoothly; but now that I am trying to run it with my own videos, I am facing a problem because the arena that I need is a Square, so wen I try to run

my_deepof_project_raw = deepof.data.Project(
    project_path=os.path.join("/data/dlc_cin"),
    video_path=os.path.join("/data/dlc_cin/Videos/"),
    table_path=os.path.join("/data/dlc_cin/Tables/"),
    project_name="deepof_cin",
    arena="polygonal-autodetect",
    animal_ids= [""],
    table_format="h5",
    video_format=".mp4",
    exclude_bodyparts=["Tail_1", "Tail_2", "Tail_tip"],
    video_scale=380,
    iterative_imputation="partial",  # "full",
    smooth_alpha=1,
    exp_conditions=None,
)

My Kernel crashes, the same problem happens when I try to run the code with:

polygonal-manual circular-manual

Also another thing that I am a little bit insecure about is what should i assing to animal_ids if i only have one subject.

I was wondering if maybe you have any suggestion of what might be happening, also I am running the code on docker, docker wise it have 46 Gb of Ram and CPU 3900x. In case you need them the videos and tables are attached Tables and Videos.zip

[NoCreativeIdeaForGoodUserName]

Hi Steven,

thank you for using deepof. Regarding the animal_ids, you simply can skip providing this input argument, if you only have a single animal (like I did in the example below). In my short test just giving animal_ids= [""] also worked.

Using your Tables and Videos with the configuration below and deepof 0.7.1 and 0.7.2 (the latter is our most recent version that should be released later today or tomorrow) I was able to create your project with your tables and videos in my own folder structure:

base_path="D:/BugReprods/Projects/Not_working_h5/"
project_path=base_path
video_path=base_path+"Videos"
table_path=base_path+"Tables"

my_deepof_project_raw = deepof.data.Project(
    project_path=project_path,  #replace these with your paths
    video_path=video_path,
    table_path=table_path,
    project_name="test_project1",  #replace this with your project name
    arena="polygonal-manual",
    table_format="h5",
    video_format=".mp4",
    exclude_bodyparts=["Tail_1", "Tail_2", "Tail_tip"],
    video_scale=380,
    iterative_imputation="partial"
    smooth_alpha=1,
    exp_conditions=None,
)

Note that video_path and table_path are intended to lead to the folders that hold your input videos and tables respectively (not the folders that will be created to hold edited tables and videos). project_path leads to the folder in which later the project folder under the name given in project_name will be created. If your project creation still does not work after using my template with your paths and updating deepof to the most recent version, you could try to use full paths instead of the partial ones you provided (like I did in this example).

For this project, I would recommend to use the polygonal-manual option since the arena is pretty shadowy. I'll let you know when deepof 0.7.2 is out since this version contains a patch that fixes a binning bug that is present in 0.7.1 (this bug leads to binning requests for the plots getting ignored. Not relevant right now, but may be annoying when you get to the analysis part).

Finally: I was not able to replicate your issue of the (I assume) Jupyter kernel crashing. If the measures above do not fix your error, could you provide me with the error report and if you are using a Windows?

I hope this helped,

Best regards, Patrick

[StevenM711]

Hello!!

Thanl you so much for your help, just tried to run it on my personal computer and it worked, I believe it might have been the other computer I was using that made the Kernel Crash, but i works now.

Also thanks for the suggestion of polygonal-manual, it worked smoothly

Regards, Steven Mora

[NoCreativeIdeaForGoodUserName]

Great to hear that I could help you. You already mentioned that the computer you tried to run it on before had sufficient RAM. Were there any other oddities that could explain why the kenrel may have crshed that come to your mind? (e.g. an older version of Windows or another specific operating system, an old version of Python (deepof is mainly designed to work with 3.9 and 3.10), or anything similar?) I am asking to keep track of potential issues that we might need to address in the future.

[StevenM711]

Hello!

When the Kernel crashed, it didn't provide any error code to indicate what was going on. This only happened when I was running my code attached to the container. It didn't occur with circular-autodetect, but it did with all the other options. I then installed deepof using !pip install deepof (if I remember correctly) in a new environment, so I wouldn't be running the code in the container. It worked on my personal computer (Windows 11, 48GB RAM, 3800X, 3060Ti), but not on the computer I'm using at work (Ubuntu 22.04.4 LTS, 64GB RAM, 3900X, NVA8, Python 3.10.12). On the work computer, when I select polygonal-manual, it doesn't display the window to select the area

[NoCreativeIdeaForGoodUserName]

Thank you for providing this information. I will run some tests on our Ubuntu systems here to see if I can get a similar error to occur (but this could take a while since you unfortunately do not get any error reports that could help me to really narrow down the reason for this crash). Often, tehse kidn of problems are related to a very specific interaction between your operating system, some of the packages, and your Python version. These kind of issues may go away "by themselves" as packages get updated in future releases. Hence, I will first do some intial tests. If I can reproduce the issue I will fix it. If I cannot reproduce it, I will wait if I get more similar reports that may provide me with more information.

I will keep this issue open for a few weeks. If you should notice that deepof starts to work again on your working PC with a new version of deepof you can let me know and I will close this issue. If I cannot reproduce it and do not get any similar reports it may be something that is extremely specific to your PC. Since you already found a workaround I will then close the issue after a while.