Core technologies we use

We harness a wide range of modern AI/ML technologies, optimized for both performance and deployment flexibility:

ML Frameworks and formats

• PyTorch, TensorFlow, TensorFlow Lite
• ONNX, RKNN Toolkit for Rockchip NPUs

Computer Vision

• OpenCV, YOLOv5, YOLOv8, MobileNet-SSD
• Custom model training for aerial and thermal imagery
• Real-time multi-object tracking and classification

Machine learning techniques

• CNNs, RNNs, SVMs, Q-Learning, Reinforcement Learning
• Kalman Filtering, SLAM
• SGD and Adam Optimizers
• Data processing with Pandas, SciPy

Embedded & edge optimization

We build solutions that are tailored for lightweight and energy-efficient edge devices:

• Raspberry Pi
• Orange Pi 5
• NVIDIA Jetson Nano
• Rockchip SoCs (RK1808, RK3588, RK3576)
• NPUs and accelerators with CUDA, OpenCL, RKNPU

We optimize models for:

• Real-time performance
• Low-power operation
• Compact memory footprint

Integration with drone ecosystems

SkyCreatures engineers specialize in integrating AI with drone flight control systems, ensuring a seamless interface between autonomous decision-making and physical flight behavior.

We support:

• MavLink protocol
• ArduPilot stack
• Real-time telemetry feedback
• Autonomous waypoint navigation
• Visual obstacle avoidance
• Smart flight behavior driven by AI inference

Development languages & environments

We support development in:

• Python – for rapid prototyping, training, and edge inference
• C++ – for high-performance deployments and embedded systems

Our solutions are adaptable across a wide range of hardware architectures and are containerized or modularized for easy deployment, update, and scalability.