Face Detection Algorithms Comparison
A comparative computer-vision assignment implementing and evaluating Haar Cascade, HOG + SVM and ResNet-SSD approaches using Python and OpenCV.
Problem
Face detection performance changes with pose, scale, lighting, occlusion and background clutter, making algorithm selection important for real applications.
My Role
Implemented and compared classical and deep-learning detector workflows, prepared the report, documented the dataset and discussed accuracy and speed trade-offs.
Tools
- Python
- OpenCV
- dlib
- Haar Cascade
- HOG + SVM
- ResNet-SSD / OpenCV DNN
- WIDER FACE dataset
Dataset / Inputs
A diverse subset of the WIDER FACE benchmark dataset in YOLO format, evaluated with precision, recall, F1-score and runtime discussion.
Results
The report highlights Haar for fast CPU-based detection, HOG + SVM as a robust classical baseline, and ResNet-SSD for stronger behavior under varied conditions.
Challenges
Keeping the comparison fair across models with different assumptions, dependencies and computational requirements.
Demo & Report
GitHub
Repository status: Private / available upon request. The public GitHub profile is linked below. For a walkthrough, code discussion, screenshots or a demo, contact me directly on WhatsApp.