yoterel / storm-net Goto Github PK

Running the training through the gui with gpu_id = 0 (only 1 cuda enabled gpu) provides the output in terminal:

Running the same training through the command-line provides the following output in terminal:

Both are snippets of the fulloutput.

The second finishes the training with nowhere near the required val loss, whilst the first hangs at epoch 49 and progresses no further.

Hi,

we managed to install Stormnet, however, we run into the following problem: for the Offline Step, rendering the synthetic data (with all default settings) runs and creates a cache/synth_data folder full with json files. However, these files do not seem to be correct. Under "valid stickers" they are always filled with "false". The training process also does not work with these data.

For reference I attached one of the json files. We don´t know what causes the problem and we are happy for any help!

image_000005.json

Need to add it uses the NFRI_CB17 template from the NFRI_R17 database for the world-> MNI coordinate transformation performed in the GUI

I have managed to get the STORM-Net GUI running. However, when I try to render synthetic data, I don’t see any progress after 10-ish minutes of rendering, even though I did check the "Monitor Progress" box (I just see the green box moving from left to right). The same goes for rendering the example model data.

When I close the GUI (while still trying to render the synthetic data), I do get the error “TypeError: cannot unpack non-iterable NoneType object”. With Traceback: “ File “main.py”, line 89, in
sticker_locations, video_names = video.process_video(args)

I was wondering if you have any clue what could be going on? Might this has something to do where all the files are stored or something? I guess the installation of all dependencies worked, since I managed to get the GUI started using Anaconda.
Or does rendering synthetic data take a huge amount of time?

Hi,

We are making a lot of progress on using Storm, with our next step being training our own model. I was wondering if you could provide a bit of guidance on how to go about doing this. I have checked the GitHub page, but I am not quite sure which files I am supposed to be uploading to which spot in the GUI. Below I have listed what the GUI requested - could you please confirm/let me know which files it’s asking for for each?

Thank you!

RENDER SYNTHETIC DATA

Template Model File: ? (what should I put here?)
Renderer: ? (Not sure what the renderer is)
Synthesized Data Output Folder: Is this just a folder we want to specify for where the output should go?
Renderer log file: Not sure what this is
Number of iterations: 100000

TRAIN STORM-NET

Model Name: Just name it?
Synthesized Data Output Folder: Is this just a folder we want to specify for where the output should go?
Training Log File: Is this something we obtain during the rendering step?

Hello,
I'm running this program on Windows and encountered this error with default settings. For the template model file, I used the adultTemplate.txt and also tried other model files but none of them worked. Here are the error messages and I wonder if you have any clues regarding this issue?

Exception in Tkinter callback
Traceback (most recent call last):
File "C:\Users\14993\miniconda3\envs\env\lib\tkinter_init_.py", line 1705, in call
return self.func(*args)
File "C:\Users\14993\Documents\STORM-Net-master\CapCalibrator\video_annotator.py", line 811, in render
elif self.is_renderer_active():
File "C:\Users\14993\Documents\STORM-Net-master\CapCalibrator\video_annotator.py", line 790, in is_renderer_active
return file_io.is_process_active(self.renderer_executable.name)
File "C:\Users\14993\Documents\STORM-Net-master\CapCalibrator\file_io.py", line 176, in is_process_active
output = subprocess.check_output(call).decode()
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xd0 in position 0: invalid continuation byte

Many thanks!

Approx 20s per json, with 100,000 iterations = 20 days to load.

Python version 3.7.16
Windows 10

This was requested so that users can input arbitrary names instead of numbers from 0 to n-1.
Useful, but requires careful changes across most of the code.

OSError: [WinError 126] The specified module could not be found. Error loading "C:\Users\User\miniconda3\envs\env\lib\site-packages\torch\lib\shm.dll" or one of its dependencies.