Building Synthetic Simulated Environments for Configuring and Training Multi-camera Systems for Surveillance Applications

• DOI: 10.5220/0010232400800091
• Publication Date: 2021-02-17T17:50:50Z
• Authors: Nerea Aranjuelo; Jorge García; Luis Unzueta; Sara García; Unai Elordi; Oihana Otaegui
• Keywords: Deep Neural Networks; Horizon 2020; Clean Sky 2 Joint Undertaking; object detection; Video Surveillance; 02 engineering and technology; synthetic data; simulated environments; Synthetic Data; Simulated Environments; deep neural networks; Object Detection; 0202 electrical engineering, electronic engineering, information engineering; video surveillance; European Union (EU)

Synthetic simulated environments are gaining popularity in the Deep Learning Era, as they can alleviate effort and cost of two critical tasks to build multi-camera systems for surveillance applications: setting up camera system cover use cases generating labeled dataset train required Neural Networks (DNNs).However, there no ready solve them all kind scenarios cases.Typically, 'ad hoc' built, which cannot be easily applied other contexts.In this work we present a methodology synthetic with sufficient generality usable different contexts, little effort.Our tackles challenges appropriate parameterization scene configurations, strategies generate randomly wide balanced range situations interest training DNNs data, quick image capturing from virtual cameras considering rendering bottlenecks.We show practical implementation example detection incorrectly placed luggage aircraft cabins, including qualitative quantitative analysis data generation process its influence DNN training, modifications adapt it contexts.