Title: Immersive Marathon Runner Application
When (time-plan): June to December 2019 (6 months)
IMRA will experiment with crowd analytics applied in the coverage of marathon and other running events. It will utilize and mix user generated content with traditional production content, if available, to provide an immersive experience to the spectators of running events conducted within urban areas. The multiple viewpoints that observe the event will provide continuous feeds of the spectators that will be analyzed in order to track runners and offer their positions as well as other interesting moments from the race in real time.
IMRA will exploit the edge computing availability of future 5G networks to offer a low-latency service in crowded areas via a white label mobile application. The app will both capture user-generated content to forward it to the backend service, as well as configure the reception of monitoring data on a per-user basis. All of the service’s components will be deployed at the edge, capitalizing on the localized nature of the event and thus, the limited spatial extend that the service will be delivered at. The crowd analytics backend will be deployed as an in-network application function and will utilize state-of-the-art research in person detection & tracking, video summarization and optical character recognition in order to identify runners and/or match their running identification numbers and create live short videoclips for each runner. A resolver-load balancer and storage components will support it, being responsible for routing incoming video feeds and serve as the connection point to the mobile apps, respectively.
Overall, IMRA will experiment with edge deployment and verify a new immersive experience service that will consume and deliver personalised and localised media. The aforementioned concept is related to a business opportunity associated with running events and is expected to boost our capacity to deliver and commercialize it upon 5G network availability.
IMRA has set the following objectives that will be addressed and measured through measurable KPIs:
- O1: To deploy and test an Immersive Marathon Runner Application (IMRA) to analyse multiple camera feeds in crowded areas
- O2: To use the FLAME infrastructure for experimenting with the offloading of computation in the FLAME edge.
- O3: To enable the preparation of D-cube offering in Immersive sports events upon availability of the 5G networks.
- O4: Reuse the FLAME offerings and provide feedback on the FLAME consortium
D-cube’s expertise on deep learning (DL) technologies and architectures will be exploited to deploy several DL models to test different configurations of the FLAME infrastructure. The foreseen DL components that will be used in the IMRA experiment are the following:
- Person detector component
- OCR for runners’ number identifier
- Crowd counting and classification component
- Video summarisation per runner or per group of runners.
Check the IMRA blog post here.