C++ Edge SDK
The high-performance embedded implementation of the Spatial Fabric. A dependency-free, header-only library designed to enable sub-millisecond safety triggers, IoT ambient monitoring, and real-time XR physical inputs on microcontrollers.
git clone https://github.com/Willow-Dynamics/willow-cpp-sdk.git
Ambient Intelligence on a Microchip
True spatial computing and robotic safety cannot rely on round-trip network calls to cloud neural networks. The C++ Edge SDK utilizes Int8 Quantized Models—microscopic payloads of intelligence generated by the Cloud Oracle. By streaming local sensor data through our zero-allocation Continuous Dynamic Time Warping (DTW) algorithms, your bare-metal devices can physically "understand" human context locally. This enables instant hardware-level feedback loops, turning basic hardware into ambiently intelligent infrastructure.
Real-Time Safety Triggers
Execute sub-millisecond reactions to specific human poses, ergonomic faults, or safety violations. Trigger machine shutdowns instantly without waiting for network-based computer vision analysis.
Low-Latency Spatial Inputs
Provide the deterministic backbone for AR/VR applications and physical gamification. Turn complex, fast-twitch physical gestures into instantaneous digital controller states.
Header-Only Bare Metal
Integrate Willow intelligence into any C++14+ project. No complex build systems or external OS required. Uses zero dynamic heap allocation to completely eliminate memory fragmentation on SOCs.
Mathematically Deterministic
Our O(N) spatial alignment kernels are optimized for ARM, x86, and RISC-V architectures, ensuring bit-for-bit mathematical parity with the Cloud Oracle across diverse hardware environments.
Int8 Quantization
The SDK natively decodes V4.0 quantized binaries. These microscopic model footprints allow you to embed hundreds of complex contextual spatial states directly onto a single, low-power chip.
Sensor-Agnostic Ingestion
Decoupled from specific camera hardware. Whether your raw coordinates come from Azure Kinect, LiDAR, Apple Vision Pro, or basic IMU suites, the engine standardizes and processes the data instantly.
#include "willow.hpp"
#include "hardware_io.hpp"
void main() {
// 1. Provision Model (RAM-Only / Zero Heap Allocation)
std::vector<uint8_t> buffer = fetch_from_firmware_storage();
willow::Model model = willow::Model::load_from_memory(buffer);
// 2. Initialize Engine & Result Container
willow::Detector detector(model);
willow::DetectionResult result;
while (true) {
// 3. Ingest Sensor-Agnostic 3D Coordinates
willow::Skeleton frame = hardware_io::get_75_point_skeleton();
// 4. Sub-Millisecond Evaluation (O(N) time complexity)
if (detector.push_frame(frame, result)) {
std::cout << "Action Detected! Confidence: " << result.confidence << "\n";
// 5. Trigger immediate hardware state change
hardware_io::trigger_safety_relay();
}
}
}