Complete code for training with zero annotations (Issue #43)

[jrbastien]

Hi,

I have incorporated NiklasWilson code and also corrected utils.py in order to be able to be able to use the lines with no bounding boxes in data_train.txt.

There was a message "Ignoreable error: Folder/File could not be accessed, may not exist ->" that I thought could mislead the user so I improved it to say "Sometimes it is desirable that your training dataset include images with non-labeled objects that you do not want to detect (i.e. negative samples without bounding box). Non-annotated images can be placed in -> /home/username/Src/TrainYourOwnYOLO/Data/Source_Images/no_object. However, this is not mandatory, you can now continue with the training."

I think that it is complete and ready to use. Feel free to merge into the master if you like.

[AntonMu]

Hi @jrbastien

Thank you for all your work! I appreciate it.

I would like to merge it at some point but I would need to review it carefully. It contains a lot of code changes that need to be merged safely.

To get things moving, please have a look at the contributing doc. It contains a list of instructions to make a successful pull request.

If your PR contains independent changes please consider splitting them up into smaller PRs.

Thank you!

[AntonMu]

Closing this due to inactivity.