Back to projects
FeaturedCompleted
Age Group Classification Pipeline: From Feature Engineering to CNNs
A comparative study and implementation of multiple computer vision architectures for 4-class facial age detection, ranging from traditional handcrafted feature pipelines to deep learning.
PythonNumPyPyTorchOpenCVScikit-learn (SVM/MLP)
Developed an end-to-end classification pipeline for categorizing facial images into four distinct age groups. The project served as a comparative analysis of feature extraction techniques, evaluating the efficacy of traditional methods (HOG and SIFT with Bag-of-Visual-Words) against modern deep learning architectures (MLP and PyTorch-based CNNs). The study addressed challenges in real-world imagery, including variations in lighting, pose, and facial alignment, by implementing robust preprocessing stages.