The observer who selects engaging scenes is a vital part of Esports.

The paper proposes an automatic observer model based on human observational data.

Our proposed method focuses on a spatial area the spectator watches.

The proposed model applied an object detection mechanism to find the spatial area.

The proposed model has higher performance than the existing event-based method.

Technologies used:

• Docker
• Pandas
• Matplot
• C++

Offline reinforcement learning is an algorithm that makes optimal decisions on a data set collected without exploration. In this paper, data augmentation is applied to the MDP structure of offline RL to improve the performance of the agent in the Atari Game.

Technologies used:

• Docker
• Amazone EC2 (Cloud Compute Service)
• Python

In this work, we propose a simple testing method, Determining the Possibility of Transfer Learning (DPTL), to determine the transferability of models based on Grad-CAM visualization of the CNN layer from the source model.

Explainable AI (XAI) is a visualization tool that shows the decision-making process in a way that humans can understand. In this paper, we propose to use Grad-CAM, one of the XAI techniques, when we visualize the behaviors of AI players trained by deep reinforcement learning (RL). Our experimental results show which part of the input state is focused on when one well-trained agent takes action.