Instant video scene analysis
Our IVD runs customized apps for different events' recognition. They combine the results of different AI algorithms and this allows them to react to the scene on the field of view. It can recognize from a combination of objects to human behaviour and interaction. The IVDs can also form a P2P network and create a multi-camera field of view.
Everything happens on the edge, so that video is not shared outside the component. The ultimate privacy tool without compromising on-scene analysis.
Automatic image segmentation pipeline
We have developed an automated segmentation pipeline that allows us to get good training datasets on any object without any human segmentation. This reduces tenfold the amount of time and cost to obtain a DNN segmentation model that can then be used at an industrial level. Our training pipeline uses automated DNN trimming (architecture and filters) to achieve the best segmentation results possible. We also adapt our system to deliver to a customized UI or we build an app so that it can be accessible from a mobile device.
Business data gathering and analysis
Our synchronized IVDs can exchange information seamlessly and cover a big surface retail area or public space. The sensors can collect and process data onboard about people flow, space occupation, client physical data, groups, occupation hotspots, people's attention span, etc. All that data is collected and processed in the IVD so no footage is shared outside the system. This eliminates any privacy rule breaking and allows the system to provide analysis as it happens. The analysis is sent to either your system, our user interface or directly to our mobile app.
You can also sell through our portal all the analysis results to major brands or rent directly part of your space to those same major brands based on our hotspot mapping. The process is automated based on retail surface structure and client attention parameters accessible through our portal. The brands pre-pay for the space and you only have to fulfill the product physical distribution orders.
Synthetic data set development
State of the art synthetic pipeline to create hyperrealistic synthetic elements for deep learning training, using semantically selected backgrounds and adapting to surfaces perspectives, lighting and texture.