hallerpatrick / ai-birds-levelganerator

Hi 🤗, I'm Patrick Haller
I am a PhD Student and currently deep diving into the field of Natural Language Processing.
In the mean time, I am looking to contribute to great Open Source projects.

~ Profile · website · linkedin ~

ai-birds-levelganerator's Issues

Build Level XML -> Json Parser

Bauen eines Parser, der die Level XML in ein Json convertieren kann.

Server & Client log parser

Writing a script that can read all logs of the server and client to extract wether a level could be played or not (in first stage).

This can heuristicaly decided if the client repeatedly is asking for new screenshots and is not making any further decision.

Setup GUI for testautomation

Setting up a gui (prefferably in javafx) for the automation task.

1. Able to set all path variables of needed programs
2. Show little interface and output
3. Profit

Manipulate the raw level images to show the contures before training more clear

Get conture of GAN generated images

The conture detector should recognize the objects of a generated image. From the contures calculate the centroid to use as reference point for the xml.

Fix converting of coordinates from Pythons pillow to cartesian xml coordinates

Setup test automation

1. After generating the xml files. those should be imported into the unity game and played.
2. From there, the game should be started aswell as the server and agent.
3. Read out the game results and do further work (more issues incoming)

Setup working agent with unity

ERROR: expecting 840-by-480 image

Setting up agent play levels

Build xml from detected contures

the xml_generator shall be able to construct a xml file in the required format. For this the conturs are parsed into game objects and the coordinates are converterted to cartesian system.

Setup a reevaluation system

Generated levels should be test (gravity at first) to determine if level is valid. For this all images (of all objects) and their generated xml should be saved and after the tests saved in a new input directory, that can be used to train the models again.

Build Level JSON -> XML Parser

Genaus sollen Level JSON zu XMLs convertiert werden können.

ySetup a solid xml generator that uses the parameter.txt properly

The generator should read the information of the parameters.txt and should succesfully build the amount of levels with the passed restrictions.

Build complete raw levels from XML

In this task every all the information, concerning the objects of a level, should be put into the raw generated image.
For now 3 different objects are differentiated with colors:

1. Pigs: pink rgb(255,105,180, 1)
2. Platforms: white rgb(255,255,255, 1)
3. Blocks: green rgb(0.255.0)
4. TNT: red rgb(255,0,0)

The chosen img format is 842x482 pixel

Report erstellen