Vision-Based Chess Detection for a Robotic Companion

Many countries are facing the aging of their population, and many elders are willing to stay at home instead of nursing homes. The elders could utilize the robots to cook, monitor the condition of health, even give them emotional companionship. Therefore, a robot that brings new solutions to the elderly care problem by accompanying elders playing chess is proposed in this paper. The robot consists of vision system, chess engine and manipulator, and the vision system is mainly discussed in this paper. The EfficientDet with anchor boxes set by k-means clustering algorithm is applied to build the vision system, which had 79.0% of mAP@0.5 in the test set. In the end, we discuss the directions of improvement.