FengJi2021/IAS_Naviprediction

IAS NaviPrediction

Pipeline Draft

Date augmentation with GAN (Generative adversarial networks)

A deep learning approach was used to complement the data set. for this approach the method Generative adversarial networks was chosen. to use the script, the file gan.py should be executed. The arguments are as follows:

• --image_path: path of dataset.
  • for generating the maps, there is in the folder data_augmentation_GAN, a folder with dataset. you can use this as a dataset or use your own maps in .png format. all maps (images) should be in one folder.
• --output_path: path of the output folder.

Example:

python data_augmentation_GAN/gan.py --image_path /path/to/dataset --output_path /path/to/output

Filter of maps

Second step is to filter the generated maps. to filter the maps, the script filter.py should be used. The arguments are as follows:

• --image_path: path of generated images.
• --output_path: path of the output folder.

Example:

python data_augmentation_GAN/filter.py --image_path /path/to/generated_images --output_path /path/to/output

Create yaml file

For the world complexity metrics, the maps should still have yaml files. to create yaml files, you should run the script create_yaml.py. he arguments are as follows:

• --image_path: path of images.

Example:

python create_yaml.py --image_path /path/to/folder/of/images

calc_ComplexityValues.py

This script calculates the complexity values of the given maps in a folder and writes them into a csv-file as input for the DNN.

Example:

python3 calc_ComplexityValues.py folderpath

World complexity

in order to calculate the metrics of worlds, script world.complexity.py should be executed. the arguments are as follows:

• --image_path: path of the image.
• --yaml_path: path of the map.yaml file.
• --dest_path: path of the output folder.
• --folders_path: if each maps and associated yaml file are stored in a folder and separated from other files, you can use this argument. this way all folders will be loaded automatically and complexity metrics will be calculated.
  • if this argument is used, other arguments must not be specified

the name of yaml file must be map.yaml, if your file has another name, please rename it to map.yaml

Example:

python world_complexity.py --image_path /path/to/folder/of/images --yaml_path path/to/map.yaml --dest_path path/to/output/folder

or

python world_complexity.py --folders_path /path/to/folders/of/maps

Step 1

1. Install world_complexity.py script

Step 2

1. Please read those notes
2. install and validate the package

Step 3

1. Test script world_complexity.py (what are the limits of the script? what could be improved?)
2. Create enough maps (~100 - 200)
3. Collect values from maps via world_complexity.py
4. Test, if the collected values can be added to one value or if a CNN is necessary