To collaborate with the local authorities in the organisation of beaches and access to touristic sites I build different pipelines that counts people or density occupation in the beaches. I also counted anchored boats and tied up in harbours. Images came from CCTV cameras and computation on the edge was necessary to avoid privacy violation.
Apart of intense use of CV2 and classical computer vision algorithms I used already mobilNet pretrained models for people and boats detection and exported to tfLite architectures to do inference in CPUs.
Occupancy and density analysis software build with openCV has been installed in three beaches serving information of free area of sands to mobiles apps and street panels in real time.
The success of those projects made headlines in the local media
Proyecto ION BEACH: control de aforo en playas para un turismo más seguro
With deeplearning I build a software to count access of people and also it counts anchored boats from a 360º moved with PTZ the authorities had access to the data in real time to regulate access to an island.
Technology stack Python + detectron2 + OpenCV + Kafka + PTZ cameras management.
