Learning system and method for automatic observing service

A system for evaluating a trained artificial neural network model, the system comprising:

• a game data processing portion that generates game input data by abstracting a game screen of a frame configuring a game video file for evaluation;
• an observing data processing portion that collects observation area information each selected by at least one human on the game screen;
• an artificial neural network model that trains to predict an observing viewport on the game screen by training to predict an area of human interest among the game screens based on a plurality of masked human data derived from the game input data for evaluation and the presently collected game input data for evaluation;
• and
• an evaluation portion that evaluates the artificial neural network model trained based on the predicted observing viewport and observation area information each selected by one or more humans.

A learning method for an automatic observing service, the method comprising:

• generating game input data by abstracting a game screen of a frame configuring a game video file for learning;
• generating a plurality of masked human data based on observation areas each selected by a plurality of humans on the game screen;
• and
• training an artificial neural network model to predict an observing viewport, which is an area of human interest among the game screens, based on the game input data and the plurality of masked human data.