ComBox Monster Vision is a video analytics platform that allows to use neural networks based on Intel OpenVINO and can be integrated into existing solutions. At the system input a video stream is received in the RTSP format, it is cut into frames in accordance with the set frame divider of the frame rate and then the frame is transmitted to the neural network selected for the camera.

For most commercial tasks and pilot projects, the adaptation of the model_zoo networks is enough. The advantage of the platform is the possibility to use of a variety of neural networks sequentially and in parallel, for example, for face detection and further construction of the sex and age structure or vehicle detection and vehicle classification by make and model.

Trajectory analysis of objects in the frame with their re-identification after overlapping is used for generating unique events. The platform is created for object detection and recognition and generating reports based on the customer’s criteria.

Fields of application: visitor counters in closed shops and at free-standing stands; counters and classifiers of vehicles; passenger traffic meters for public transport; tracking and re-identification of people in common areas; building the age and sex structure of visitors; building a heat map by area; safety of detached objects; recognizing faces and emotions of people.