This repository demonstrates how you could potentially use a convolutional neural network classifier to determine game save checkpoints from a video game screenshot, allow screenshots and photo mode to be used as save files. This was a fun technical experiement but is ultimately impractical for a number of reasons.
For more on why it's impractical and a video walkthrough of this project, as well as the alternative approaches that were taken including steganography, EXIF tags and QR codes, check out my YouTube video on the subject here.
- Place your .mp4 video files in frame-extraction/videos/
- Build the frame extraction image:
docker build -t frame-extraction frame-extraction/
- Run frame extraction:
docker run -it -v $(pwd)/frame-extraction:/data frame-extraction
- Retrieve the split video frames from frame-extraction/frames/, and place in the appropriate training-images/ subdirectory such that each subdirectory represents a class you want to train/predict.
- Build the training image:
docker build -t tf-checkpoint .
- Run tarining:
docker run -it tf-checkpoint