It was another successful ITS World Congress for Cloud LSVA. The EU-funded project on Large Scale Video Annotation for autonomous vehicles has showcased its latest developments in many different occasions in the world’s largest event in the ITS industry.

The first presentation was held by ERTICO’s senior manager Erwin Vermassen, who presented the paper “Cloud Based Large Scale Video Annotations to improve mapping and mobility for connected, cooperative and automated transport” in the well-attended session called “The future evolution of ITS”. This paper outlined the technical challenges faced by autonomous vehicles when trying to gain a detailed view of the surrounding environment, such as unforeseen obstacles (like pedestrians, lane restrictions or work areas). A fast, almost real-time annotation of video data, produced by the vehicle on-board cameras and other sensors is therefore imperative to realize the vision of automated transport in a near future. In this occasion, the Cloud LSVA consortium presented the project’s solutions, which consist in using deep learning techniques to realize the semi-automated identification of objects in a scene, the events and relation between the objects and their environment.

 

Oihana Otaegui, Head of Intelligent Transport Systems and Engineering Department in Vicomtech, also presented Cloud LSVA’s solutions at ERTICO’s stand. Oihana focussed on Cloud LSVA’s solutions to record large amounts of video data by the vehicle, which consist in shifting from manually automatically annotating data, considering an intermediate stage of manual and semi-automated annotation along the process. This way, corrections can be made, achieving more accuracy.

 

In the framework on semi-video annotation, Marcos Nieto, Principal Researcher at Vicomtech at the Department of Intelligent Transportation Systems and Industry, and Cloud LSVA consortium, spoke about “Video Anonymization VIA Semi-automated annotation” at the Special Interest Session “Automated vehicle data sharing enabled by Feature Extraction and Anonymization”. Modern test cars are recording massive amounts of video footage on streets, roads, and in many other contexts. Images of faces, license plates, bodies, telephone numbers are being recorded and this raises the issue of privacy. In his presentation, Marcos focussed on a possible solution to the problem by using semi-automated processing platforms to anonymize videos and video encryptation standards to ensure video original is kept/recovered.

The Cloud LSVA consortia is now preparing for the final steps of the project. Join us on 29 November for the final event in Stuttgart, Germany (venue to be announced soon). This event is free of charge and open to the public.

Click here to view the agenda.

Click here to register.