Trained Real-Time Face Mask Detection System with YOLOv4

Trained and evaluated a YOLOv4 deep learning object detection pipeline in Python using OpenCV and Darknet/PyTorch to recognize face mask compliance in real-time camera video streams with high frame rates.

Implemented custom dataset preprocessing, bounding box annotation parsing, and inference pipelines for multi-face detection.

Dataset augmentation (rotation, illumination shifts, occlusion) and careful anchor box tuning are crucial for computer vision models to generalize across varied real-world camera angles and lighting environments.

Quantize the model using ONNX / TensorRT to lower latency for real-time edge deployment on cost-effective embedded hardware.