Applied Computer Vision Suite (YOLOv8 / YOLOv11)
Real-time detection systems: from gesture control to surveillance anomaly detection.
A set of applied computer-vision pipelines: real-time gesture and object detection (YOLOv8), a security and surveillance anomaly-detection system (YOLOv11), and an ECG signal-processing pipeline extracting diagnostic features from waveform data.
The Problem
Real-time edge vision requires high frame rates and low compute footprint, while medical waveform processing demands high feature fidelity.
The Approach & Implementation
Trained custom YOLOv8 and YOLOv11 models optimized via ONNX/TensorRT for 60 FPS edge inference, paired with SciPy/OpenCV signal processing algorithms for ECG waveform analysis.
Measurable Outcome
Achieved 60 FPS real-time detection with sub-35ms frame-to-prediction latency and automated diagnostic waveform feature extraction.