cultionet

cultionet copied to clipboard

These are the changes I've made regarding the data augmentation. The change in create is what I believe is the appropriate change to fix the time series error I was...

MatthewPierson90

(bug) Prediction with window size slightly too small

4

I just encountered an error doing a prediction on a test image. It seems to be a result of the test image shape being just barley bigger than a multiple...

MatthewPierson90

This may not be cool, I have added a weighted averaging to the Norm Values (since my epochs are all on different data) It definitely has some problems associated with...

MatthewPierson90

allows for some pathing changes in the predict_image and persist_dataset functions which can be set from the cli

MatthewPierson90

[bug] ValueError: `x` must be strictly increasing sequence.

3

I run into the following error running 'cultionet create' with exactly one veg index. It can be avoided by not doing any time augmentations. I have also gotten the error...

MatthewPierson90

[bug] Possible binary incompatibility numpy==1.21.0

6

I get the following error running "cultionet create", updating Numpy fixes the error, but I am not sure if that update causes my other issues. Traceback (most recent call last):...

MatthewPierson90

[bug] EdgeDataset processed path

2

I don't think that this affects functionality, but I did come across it. When calling EdgeDataset, if I call ds = EdgeDataset(root = data_directory_path) where data_directory has the structure data_directory:...

MatthewPierson90

[bug] TypeError: __init__() got an unexpected keyword argument 'every_n_val_epochs'

1

I get the following error when running 'cultionet train'. It can be fixed by changing (Line 83) every_n_val_epochs -> every_n_epochs in model.py Traceback (most recent call last): File "/home/matthew/.pyenv/versions/cultionet2/bin/cultionet", line...

MatthewPierson90

[documentation] Information about region numbering

1

It would be useful if it were mentioned in the documentation that regions have to be 6 digits with 0's as place holders.

MatthewPierson90

[question] About training and prediction data

1

The model allows me to use data of any shape (HxW) for prediction, but for training I have had issues trying to train the model on data of different dimensions....

MatthewPierson90