MENU
Contact Us




Project

Work time recording based on Facial Recognition

Recording of entrance and exit times for employees



CPU

ARM Cortex-A53

Manufacturer

Toradex

OS

Torizon

Skills

Pytorch
Docker
MongoDB

Project size:

Difficulty:




Request/problem:

The customer wanted a prototype to evaluate the possibility of transitioning from a classical timekeeping system to a camera-based system with facial recognition for recording worktimes. The solution must be capable of distinguishing known employees from strangers and should be a single solution for data acquisition and visualization.

Solution:

Pre-trained models for face recognition are available as open-source, and together with MongoDB and dashboarding solutions the whole system can be implemented on an embedded device. Since the number of employees is fairly limited, memory and storage are sufficient to support this application. The device needs an HMI to give feedback to the user whether recognition was successful and must support a webcam for capturing video.

Architecture:

Results:

The hardware consists of the Toradex Verdin Mini SoM, which is connected to an HDMI monitor that displays the video feed of a webcam connected to the SoM via USB. When standing in front of the webcam, an employee can clock in, which starts the recognition process. At first, the face is detected via MTCNN, and the resulting region is cropped and analyzed by a pre-trained FaceNet model. The resulting 128-dimensional embedding is compared to the embeddings of known employees and if a match is found, the entrance or exit time is written to a database. Recorded durations can be viewed on a dashboard that is available in the local network.

Screenshots:

Ressources: