Version and Store training datasets

[DylanG5]

To have reproducible model results it is important to organize and save datasets that are being used for training. To this end there are a few solutions we can employ:

1. In this repo or in another one store a dataset in an organized manner as a zip file
   • requires zip file to fit inside git repo (2-4 gb max per file)
2. Utilize a cloud storage solution such as Amazon S3
   • no restrictions on file size, easy to upload / download
   • not free, though very inexpensive overall
3. ...

Success Criteria:

• A script that provides a simple way to download or upload versions of a dataset and abstracts the storage implementation. ie script.py --version 1.0 --type walking --path /.../ would download only walking data from dataset V1.0 to a given path
• An inexpensive and effective way to place our data

The final solution can be something super simple, as long as it's fairly scalable and free / cost-effective

[DylanG5]

The goal for this issue is to lay the foundation for training arbitrary models on select sensor data. The 2 people on this ticket are going to be working on training a movement classification model, where you can predict the current type of movement (walking, running, climbing etc) to be used either along with or as a backup to the lstm model. The issue hasn't been written yet but once I get some background info and write it, I will assign the two of you.