Cart Path Recognition in a Golf Course Using Deep Fully Convolutional Networks

Gukjin Son, Junkang Kim, Wooyoung Jung, Youngduk Kim


Autonomous driving vehicles are a growing reality. Autonomous driving related industries are also growing. We believe that an unmanned golf cart is an appropriate application area for such autonomous driving. So, we have commenced the project of the golf cart the purpose of which is to provide an assistant to golf players. The golf cart includes almost all the functions the Autonomous driving should have. Image processing and recognition capability are of course one of the key functions that a golf cart needs. To develop such a kind of a golf cart would take much time. Thus, we have started the research and development of a demonstration prototype, which can perform some of the main tasks of the golf cart. In this paper, we will deal with the recognition of a cart path by themselves, trained end-to-end, pixels-to-pixels. We have collected and annotated the cart path of challenging scenes captured in a golf course. And we developed a deep learning network for a cart path. As a result, we achieve 93% accuracy and 50ms inference time.


fully convolutional network; golf cart; cart path; Semantic Segmentation; deep learning; autonomous driving

Full Text:



  • There are currently no refbacks.