Activities recognition with computer vision


As a proof of concept, we selected a complex activity:

Distance estimationPose recognitionObject detectionA human is sitting on the sofa

We run the human activity recognition using Luxonis LUX-D camera.

Main 4k CameraDual 720p 120Hz Camerasfor stereoscopy and depthMyriad X chipfor machine-learning algorithms

Most smart home systems are based on basic sensors like presence sensors, weight presence, temperature sensors. But with this kind of devices, human activity recognition is painful since it requires new sensors for each activity.

With computer vision and machine-learning, human activities recognition is more scalable.

We successfully created a proof of concept able to recognize our target activity. Besides, our implementation is robust enough to recognize other home activities and can even be customized by final users.

You want to know more about this project?