POLDER is an international project within the framework of Smart Cities and of which ACCURO is the leader.
Within the scope of this project, ACCURO has developed different algorithms to offer a new mathematical modelling service that discovers the behaviour of complex and dynamic population groups in urban environments, relating the components, entities or subjects identified and labelled within a system that have relationships with each other.
Activity detection is divided into two stages:
- In the first stage, videos have been processed to generate event proposals to locate the candidates and the activity they are carrying out spatially and temporally.
- In the second stage, characteristics were extracted and a geo-temporal classification and post-processing were carried out to generate the results of the activity detection.
Different R-CNN (regional-convolutional) deep learning neural network architectures were trained to be able to detect different objects. These neural networks were trained for different use cases related to the tourism sector and COVID-19 security measures (detection of masks, groupings of people and safety distance). In the demonstrator the result of the system will be presented.
The obtained detectors detect different objects thanks to the fact that the neural networks have been trained with a large number of images, while in the second stage the classification model aims to classify and send the detected data to a database.
Finally, the data collected from the database has been used to be interpreted in PowerBI, using intelligent functions that can provide dynamic representations for decision making.
ICT Systems, IoT, Artificial Intelligence
Entity holder of the technological demonstrator
Iván Becerro, CTO – Research and Development Director
Demonstration space Expo Industria 4.0 Burgos. Expo Zone Floor 3 Forum Evolution: Demonstrator nº 11