README

This directory contains tooling to map NeuroRace data. It is a prequisite that you have access to a directory containing such data.

NeuroRace's data packages usually contain a .yaml with metadata and a subfolder system like zed/left/image_raw_color/compressed, containing the cameras pictures.

The 2 functions this tooling contains are mapping NeuroRace data and augmenting the mapped data.

These functions have been split and can be executed separately to optimise runtimes.

The datamapper is designed to be deployed on a remote system and for a path containing one or more packages of NeuroRacer data. It will then look for the .yaml, containing the images metadata and the image data for all folders down the path.

The augmenter is then used to augment the mapped images.

Map the data

1. Check if the config.py file contains the correct target path for the NeuroRace Data as SOURCE_DIR.

2. If 1. is true, usepython3 datamapper.py to map the data. It will produce a mapped.csv at the location you started datamapper from.

Augment the data

1. Have you mapped the data?
2. If 1. is true, move the the mapped.csv file to the directory containing augmenter.py .
3. Check if config.py contains the path you want your augemented data to end at as TARGET_DIR.
4. In config.py , adjust how to many pictures you want to fill the bins up to with FILL_UP_TO.
5. Use python3 augmenter.py to augment the data. It will be output to the TARGET_DIR path you set and consist of all the images generated during the augmentation, along with a .yaml file, mimicing the original .yaml. This .yaml features the path of the augmented image along with metadata used during training of NeuroRace models.

Visualise how much data was augmented

After augmenting the data, there will be a numbers.csv in the folder you started augmenter.py from. It can be visualised using

python3 visualise.py

Known issues

visualise.py might looking for columns that are falsely labeled or missing. Adjustments can be made fairly easily by changing column names in eithernumbers.csv or visualise.py .

Apart from that, the code has no known issues. Make sure not to end paths with / since os.path.join does not like that.

Questions? Contact me: hello@tobiaskolb.dev