Shypd.aiThe largest open index of AI toolsRTNeural is an open-source library for real-time neural network inferencing, specifically designed for low-latency applications like real-time audio processing. It supports various neural network architectures and allows for dynamic or compile-time model creation.
Trending
About
RTNeural is a lightweight, open-source C++ library engineered for real-time neural network inferencing, with a strong emphasis on applications requiring low latency, particularly in real-time audio processing. It enables users to export trained neural network weights from popular Python frameworks like TensorFlow or PyTorch into a JSON format that RTNeural can then read. The library supports a range of common layers including Dense, GRU, LSTM, Conv1D, Conv2D, MaxPooling, BatchNorm1D, and BatchNorm2D, along with various activation functions such as tanh, ReLU, Sigmoid, SoftMax, ELu, and PReLU. RTNeural offers both dynamic run-time model creation and a compile-time API for enhanced performance when the model architecture is fixed. It supports multiple backends like Eigen, xsimd, or the C++ STL, allowing for optimization based on specific performance needs and target platforms. The project is actively maintained and welcomes contributions for further improvements.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
