Improper Face Detection at Frontend - Finding faces that do not meet requirements
Date
September 2023
Service
Employer
Paytm Money
Project Overview
The users had to upload a selfie to open a trading account with Paytm Money. That photo had to match their ID uploaded earlier. The main job was to prompt users that the selfie clicked did not match the requirements and that the user should upload a new image to reduce his onboarding time.
Key Highlights
To alert the users who upload improper photos of their faces during the onboarding process with mentioned guidelines regarding the quality of the photo such as brightness, contrast, face too close to boundaries, face is spoofed, user not wearing a proper shirt, etc.
Used ML models and image processing techniques for filters, reducing the photo rejection at the backend by 5%.
Used Python and FastAPI to create RESTful API for checks.
