All Tools
N
Dev ToolsFreeOpen Source
NCNN
High-performance neural network inference for mobile and embedded devices
BSD-3-Clause
ABOUT
Deploying neural network inference on mobile and embedded devices is constrained by limited compute, memory, and battery. Standard frameworks designed for desktop GPUs don't fit. NCNN is a pure C++ inference framework built from the ground up for mobile — it does not depend on third-party libraries, supports multi-core optimization and ARM NEON, and gains GPU acceleration through Vulkan. Models from PyTorch, ONNX, and TensorFlow can be converted with a single CLI tool (PNNX) and run efficiently on phones, tablets, and IoT hardware.
INSTALL
pip install ncnnINTEGRATION GUIDE
1. Deploy a real-time object detection model on a smartphone camera feed with minimal latency
2. Run an LLM on-device using the ncnn Vulkan backend for GPU acceleration
3. Port a PyTorch segmentation model to a Raspberry Pi for edge AI inference
4. Integrate face recognition into an iOS or Android app with 0 third-party dependencies
5. Batch-convert ONNX models to ncnn format and deploy across hundreds of device variants
TAGS
mobileinferenceneural-networkon-devicecross-platformarmvulkandeep-learning