Object detection with deep learning in crowded scenes

Abstract

The field of object detection within densely populated scenes is a persistent challenge that is still studied in attempts for improvement, primarily due to the difficulty in detecting overlapped objects. This project studied and discussed multiple deep learning methods in object detection that is relevant to improving this field. In particular, the application of the "One Proposal Multiple Predictions" (OPMP) method is focused on due to its promising performance shown in the crowd detection field. The findings demonstrate the viability of multiple object detectors, and OPMP, in improving detection accuracy, making it a valuable contribution to object detection technologies in crowds.