GitHub

Auto-driving game AI with TensorFlow deep learning still needs to find a way to collect more data without accessing the game API.

To extract more information is needed:

[] Edges & Contours: Detect edges using techniques like the Canny edge detector. From these edges, contours (boundaries of objects) can be extracted.
[] Hough Transforms: Detect specific shapes, like lines or circles, which can be useful for games with clear geometric elements.
[] Color-based Segmentation: If certain game elements have distinct colors, segment the image based on color ranges in color spaces like HSV.
Optical Character Recognition (OCR): Speed If the game has readable text (like scores, timers, or stats) in the screenshot, use OCR techniques to extract this textual information.
[] Semantic Segmentation
Use deep learning models to classify each pixel in the screenshot, segmenting the image into meaningful regions (e.g., road, car, sky).

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
.gitattributes		.gitattributes
README.md		README.md
controls.json		controls.json
dataMining.py		dataMining.py
dataset.csv		dataset.csv
play.py		play.py
play2.py		play2.py
play3.py		play3.py
play4.py		play4.py
play5.py		play5.py
record.py		record.py
train copy 2.ipynb		train copy 2.ipynb
train copy 3.ipynb		train copy 3.ipynb
train copy 4.ipynb		train copy 4.ipynb
train copy.ipynb		train copy.ipynb
train.ipynb		train.ipynb

Provide feedback