Generate a Region of Interest (ROI) for the highest scoring zoid orientation and location as determined by CERL models.
The all_analysis.py and translate_test.py must be placed under the Tetris-AI repo when downloaded in order to use some of the functions defined in those files.
Running the all_analysis program will prompt the user for a file containing data from a singular game. (Do not forget .tsv extension) The analysis program will create and write to 4 files:
-
[game_data_file]_predictions
Gives possible moves to be made per episode, data is formatted with the header: episode_number zoid move_row move_col move_rot move_score
-
[game_data_file]_analysis
From all possible moves per episode (as given in the _predictions file), this file will tell you whether the move actually made was an option in _predictions, and if so, its rank (# points it achieves compared to the other options). Data is formatted with header: episode_number zoid move_row move_col move_rot move_score move_rank num_options
-
[game_data_file]_best_moves_byep
From all possible moves per episode (as given in the _predictions file), this file will find the highest scoring move per episode, output it and the zoid representation. Data is formatted with header: zoid move_row move_col move_rot move_score zoid_rep
-
[game_data_file]_zoid_reps
From the _best_moves_byep file, the ideal zoid placement will be taken and used to create a file of only the zoid representations on the Tetris board. This will then be used to define the ROI within the game for every episode.