Facial Detection and Recognition Project

Facial Detection and Recognition with Age and Gender Estimation

An in-house project designed to showcase our capabilities in security solutions to clients at Wavelabs.ai. This system integrates facial detection and recognition with age and gender estimation, capable of real-time inference, even on modest hardware.

Project Highlights

• Developed a robust facial detection system using ResNet50, trained on a dataset of over 5000+ images.
• Implemented Histogram of Oriented Gradients (HOG) for face extraction achieving 40+ FPS with a resolution of 416x416.
• Utilized an ensemble approach with two separate models for age estimation (regression) and gender classification (classification).
• Achieved real-time inference on an i5 2.3 GHz processor with a frame rate of 8-10 FPS.
• Encoded facial features into a 300-dimensional vector for accurate face identification.

Tech Stack

• Computer Vision: OpenCV for image processing and face detection.
• Machine Learning: Keras for building the deep learning models for age and gender estimation.

Additional Work

Alongside the facial recognition system, I also contributed to an image classification project:

• Collected and manually labeled 3000+ images from the internet for multiple base architectures.
• Applied object detection techniques to the dataset to create bounding boxes.