Embedded AI human detection

Real-time drone-based human detection system

Developed an advanced AI-powered autonomous drone system with embedded human detection capabilities using NVIDIA Jetson Xavier NX, achieving real-time inference optimization for surveillance, search-and-rescue operations, and emergency response scenarios.

Project overview

To create an autonomous drone-based surveillance system capable of real-time human detection, I engineered a complete end-to-end solution involving custom drone assembly, AI model optimization, and embedded system integration. The project involved assembling a HolyBro drone, designing and 3D printing a custom payload module for NVIDIA Jetson Xavier NX integration, and implementing TensorRT-optimized AI models for efficient human detection during flight operations.

Achievements

Hardware Integration: Successfully assembled HolyBro drone with custom 3D-printed payload module for seamless Jetson integration.
Real-time Performance: Achieved optimized inference speeds using TensorRT optimization on resource-constrained embedded hardware.
AI Model Deployment: Implemented advanced computer vision algorithms capable of accurate human detection from aerial perspective during live flight operations.

kiosk ground station software

One-click autonomous drone mission control system

Engineered a comprehensive ground station software solution using Qt framework with integrated MAVLink communication protocol, enabling seamless drone control and autonomous mission execution. The system features an intuitive KIOSK-style interface designed for operators to initiate complex flight missions with minimal technical expertise required.

Project overview

To streamline drone operations and reduce operator complexity, I developed a complete ground station software solution that bridges the gap between sophisticated autonomous flight capabilities and user-friendly operation. The project involved designing and implementing a Qt-based desktop application with custom UI components, integrating MAVLink communication protocols for real-time drone telemetry and control, and creating an intuitive workflow that allows operators to execute pre-planned autonomous missions through a single-click interface.

Achievements

One-click Operation: Simplified complex autonomous mission execution to single-click interface for non-technical operators.
MAVLink Integration: Successfully implemented full MAVLink protocol stack for reliable drone communication and telemetry monitoring.
Custom Qt Interface: Developed intuitive KIOSK-style UI with real-time status monitoring, mission planning tools, and emergency control optimized for operational efficiency.

ai-powered change detection system

Advanced satellite imagery analysis and monitoring system

Engineered sophisticated AI models for automated change detection in satellite and aerial imagery using advanced deep learning techniques. The project involved developing custom neural network architectures, implementing model optimization strategies, and creating robust algorithms capable of identifying temporal changes in geospatial data with high precision and accuracy.

Project overview

To address the growing need for automated monitoring of environmental and urban changes, I developed a comprehensive AI-powered change detection system capable of analyzing satellite imagery over time. The project involved designing custom deep learning architectures specifically optimized for change detection tasks, implementing advanced computer vision techniques for multi-temporal image analysis, and creating robust preprocessing pipelines for handling diverse satellite imagery formats and resolutions.

Achievements

Enhanced Accuracy: Developed custom AI models that significantly improved change detection accuracy compared to traditional methods through advanced optimization techniques.
Model Optimization: Successfully implemented model compression and optimization strategies for efficient deployment while maintaining high performance metrics.
Scalable Solution: Created robust algorithms capable of processing large-scale satellite imagery datasets for real-world geospatial monitoring applications.