Best presentation award at International Conference on Intelligent Vehicles and Applications (ICIVA) 2023
Our research entitled "Event Data Representation based on Time Stamp for Pedestrian Detection” tackled with pedestrian detection for vehicle application by using event camera. The reason why we chose the event camera for this application is to make the most of its low energy consumption and high dynamic range. In event data processing, there are many difficulties such as poor pixel intensity, noises and asynchrony. For solving these issues, in addition to using the data representation shown at SSII 2023, we used patch based image processing to suppress false positive even in noisy situation. Then, we prepared various scenes in training dataset such as closed/distanced pedestrians, various postures, etc. Eventually, we could improve detection score and could achieve better performance than SOTA.