Universitas Dian Nuswantoro conducted a study on an unsupervised approach for traffic sign recognition based on a bag-of-visual-words model. The activity involved UDINUS researchers Supriyanto, Catur; Luthfiarta, Ardytha; and Zeniarja, Junta. The study addressed traffic safety through the context of Advanced Driver Assistance System, or ADAS, which is described as a system intended to help and direct drivers in order to improve traffic safety. Within this context, traffic sign recognition was treated as an important part of ADAS. Traffic signs were described as warning signs placed at the side of or above roads to provide detailed road information to drivers.
The proposed approach was unsupervised. According to the source, this means that the approach does not require labelled data or a training process for traffic sign recognition and can assist when there is a large amount of unlabelled data. The experiment used the German Traffic Sign Recognition Benchmark dataset, identified in the source as a public dataset for traffic sign recognition. The reported results indicated that a large number of visual words enabled the proposed method to produce high accuracy and good-quality clusters. However, the source also stated that the accuracy in the study remained very low. The stated reason was that each image produced only three to four keypoints, while the small image size affected the number of keypoints.
Through its focus on traffic sign recognition and traffic safety, this UDINUS activity relates to the stated goals of good health and well-being and sustainable cities and communities. The evidence specifically presents a research method, its dataset context, and its reported findings. The activity is institutionally framed as UDINUS research concerning an unsupervised computational approach for traffic sign recognition.