The motive for Pepper robot is to find a tic_tac_toe board in the environment, localise itself accordingly and then play a game of it in real time with a human.
Start the game by saying Hi to Pepper, it will prompt you all the way till the end of game after that. The detection of cells inside the board is made using YOLO with a training of around 300 images. I would suggest to have one's own training with a better dataset, but for trial one can ask for its yolo_weights from me.
I would highly recommend using python virtual environment for installing dependencies used in pepper programming. For installation of python virtual environment one can follow the guide.
pip install -r requirements.txtOne also needs to add naoqi a third party python package for using qi drivers one can find the python package here.
If having a trouble while adding the python package on mac follow this repo.
workon <virtual_env_name>
python app.pyIt is programmed in such a way that it supports the following features :
- Can play even on a rough hand made board.
- It will never loose a game If played legally. Game is based on :
min-max algo - It has heuristic learning with the help of radial enclosure of centeroids from earlier detections, for creation of board where it will try to detect board configuration even if incomplete detection occurs.
- It will try to detect, if one cheats or hasn't made a move and even if human plays computer's coin [ 'x' / 'o' ] in their move.
- Prompts to narrate the board, if human is unable to make a legal move.
-
- Video Link
- Other Video Links
MIT License 2018 © Vikram Singh and contributors