Business Context

Employed Insurance agents visit their prospective leads for educating on insurance policies and selling the right policies. For the banks and insurance agencies, it is important that their agents visit customers and build trust & long-term relationships.   Tracking visits of the agents to their client locations is important for planning and accurately capture visits.

In this project, we have built a simple and effective mobile application driven via Deep Learning Model. We will explain the flow of the app and will start first with the deep learning model.

Deep Learning Model – Development

• Labeled Data – For the list of employees, we have collected facial images. For an effective Face Recognition Model, we need to have at least 15-20 different images with different poses and variations for each of the employees.
• Used Facenet Model for creating image embedding for each of the images and then built a Support Vector Machine (SVM) for Face Recognition /Classification Model. The Model trained is saved for Prediction Pipeline
• Of course, we have taken care of some of the fundamental points around modeling such as scaling, generating/augmenting samples by rotations or changing intensity, and validating the model on different samples of images.

Django based API Framework

We have developed APIs to get the captured image of the employee, employee mobile number (primary), and GPS location (latitude and longitude) of the mobile device.

Image captured will undergo a few steps – embedding and then model is called to get the predicted details based on the image.

If the predicted mobile phone/Name is the same as input, it saves the details otherwise, it sends the response back to the mobile device /via API.