The advances in Artificial Intelligence have provided powerful tools for perception in road applications, such as Traffic Monitoring or Autonomous vehicles. However, detecting and classifying objects from top-view or aerial imagery becomes more challenging, especially in urban cities; due to the variation in size and scale of the objects; in addition to the lack of the datasets that contain information about the object classes and their 3D positions. This paper introduces a novel algorithm to refine the semantic images captured from CARLA simulator; refining the semantic images, creating a new dataset with more classes of the objects, and calculating its 3D bounding box positions. The proposed method and dataset have been validated by training a CNN - based 2D object detector, and inferring with real aerial images from VisDrone-2019, and from fixed infrastructure cameras i.e. cameras mounted on Traffic lights. The obtained results illustrates the improvements of using the proposed dataset in the object detection process, with real top-view images in complex environments.
Reducing the Breach between Simulated and Real Data for Top View Images
2021-07-11
3988343 byte
Conference paper
Electronic Resource
English
REDUCING THE BREACH BETWEEN SIMULATED AND REAL DATA FOR TOP VIEW IMAGES
British Library Conference Proceedings | 2021
|REDUCING AND PROCESSING SIMULATED AND REAL-WORLD RADAR DATA
European Patent Office | 2024
|Automotive engineering | 2015
|Overtopping and breaching of dikes – Breach profile and breach flow
HENRY – Federal Waterways Engineering and Research Institute (BAW) | 2010
|Online Contents | 1993