Colab training issue
jelrod14 opened this issue · 4 comments
This is similar to issue 46 but I am not sure how they corrected this. I am using the example file and video, and have placed them into a folder called VAME_Example in my drive. I initialize the project and move the csv into the pose folder, but the video doesn't get moved into the project videos folder despite having the correct path to the video.
videos = '/content/drive/MyDrive/VAME_Example/video-1.mp4'
This is causing (I think?) the create training dataset to fail:
`Creating training dataset...
Using robust setting to eliminate outliers! IQR factor: 4
Creating trainset from the vame.csv_to_numpy() output
ValueError Traceback (most recent call last)
in
----> 1 vame.create_trainset(config)
1 frames
/content/VAME/vame/model/create_training.py in traindata_fixed(cfg, files, testfraction, num_features, savgol_filter, check_parameter)
237 X_train.append(X_z)
238
--> 239 X = np.concatenate(X_train, axis=0).T
240
241 if savgol_filter:
<array_function internals> in concatenate(*args, **kwargs)
ValueError: need at least one array to concatenate`
Not sure why it is trying to call the csv to numpy, because I did not use this as I egocentrically aligned the data
Thanks in advance!
came into the same problem
Managed to figure it out, for some reason my project was initializing wrong and not copying video correctly. After restarting from scratch essentially and renaming all project variables it started working and copying video.
I am now running into a new problem, however, where only some of my videos are being copied into the video folder upon project initialization. I am not sure if this is a google drive issue or a VAME issue. I have about 200 videos and it is only copying the first 58
Thanks in advance for the help.
Hey,
Sounds like a google drive issue, we tested Vame project initialization with a couple of hundred videos already.
Best
Pavol