At MoogleLabs, we believe in breaking boundaries and creating products that solve real-world issues. And one big problem that almost all corporations face is keeping a track record of employees' attendance. It is an overhead task that can include several human errors.
Moreover, the pandemic rendered some of the most accepted methods of automated attendance useless, especially fingerprint scanning. So, businesses needed a new and improved way of marking attendance automatically while ensuring zero contact with a machine.
Automating the process of attendance marking can save companies significant money, reduce employees' effort, and help companies improve productivity.
So, we started working on an attendance system that does not require more effort than standing in front of a camera and created the ultimate Facial Recognition system. It also works when the employees pass the camera at a legible angle. Additionally, it is possible to train the model to accurately mark attendance or in-and-out time of employees during the office hours.
As facial recognition relies heavily on machine learning, we first focused on collecting data. It included collecting image from the employees of the company that works as the base for machine learning.
Then, we applied a machine learning algorithm to the video feed to detect & recognize the face.
The system is designed to detect the video feed and compare it to images available in the database, which also contains every employee’s details. If the computer finds a match in the database, it will mark the attendance and save time.
Capture Video Feed -> Apply Machine Learning Algorithm -> Teach System to compare video feed with database to mark attendance with time.
The technology uses the Internet of Things (Video Camera) and Machine Learning algorithm to automate the attendance system.
After the initial setup by the experts, there are provisions for the users to add and remove employee details on the dashboard.
To use daily, run the system, keep the video recorder on, and you are good to go.
Office In-and-Out Time tracking and record-keeping.
Option to keep tracking of total out-time during the working hours.
Accurate data-keeping of all attendees.
Added flexibility for admins, allowing manual attendance.
At the beginning of the process, we had limited images available for training the machine learning system.
We gathered all the employee images from the organization. However, it only allowed for face detection when the user was in front of the camera. In the case of a side view, the system had issues with detection for the individual. Hence, we used a one-shot learning to rectify the problem.
Weak internet connection hindering the process of image recognition and attendance Initially, the attendance system worked fine on a stable and strong internet connection, but during periods of weak internet, it could not take the roll call.
We created a system where the resolution of the video feed will change as per the internet connection strength to enable quick attendance check even on poor internet connection.
Compatibility Issues with IP Cameras
The application needed to be compatible with a range of IP cameras for maximum effectiveness and scalability.
The system we created needed to work with Internet Protocol Cameras for maximum utility for organizations. So, we created a code to make the system compatible with the maximum IP Cameras currently used in the company's premises.
We created an application that could mark the attendance of all employees.