The project has achieved the first phase with the fulfilment of the definition of interfaces and requirements and regulatory compliance analysis, as well as the identification and selection of suitable AI-camera technology. This is a relevant step ahead to follow up with the system design, which will include the development of the preliminary sensing camera platform design, a critical next step in the project. The current phase accomplished raises the project technology up to TRL3, which shall achieve TRL5 by the end with the cabin demonstrator.
The consortium has successfully completed the analysis of the data legal reference guide, the system and interface architecture, and the TRL plan. Furthermore, several technology scouting reports have been undertaken to select the technology that will be the baseline of the SmaCS development activity. This includes the selection of appropriate camera sensors and lenses, AI-processors, and software tools for optimal image acquisition and image content analysis with Deep Neural Network (DNN)-based recognition methods. The selection of the ideal camera for optimal image acquisition has required defining all the important and impacting parameters that affect the image quality. These include optical lens parameters (spatial resolution, deep of field and luminosity), sensor specifications (quantum convert, pixels and format), and the camera interface (connectors and protocols).
Currently, the project is analysing the sensor setup with the assistance of a simulation tool and the built cabin mock-ups. The simulation tool used in the project can resolve legal or privacy issues, which often make the use of real data hard or impossible. Both the simulation tool and the cabin mock-ups will allow generating a dataset of labelled images for training the system, using closer content to the kind of expected images for the required camera placements. Additionally, preliminary tests with pre-trained generalist DNN-based object detection models are also being analysed to assist with the training and deployment of the recognition methods.
The next phase of the project is expected to be achieved by the end of the year with the preliminary system design. So far, the Covid-19 outbreak hasn’t delayed critical milestones, although the project is closely monitoring the development of the project to analyse potential impacts on delays that might be considered depending on the new outbreaks