Empowering Edge Devices with Lightweight Neural Network Inference Framework - TFLiteBEAM

Explore the development of TFLiteBEAM, a lightweight neural network inference framework designed for edge devices, in this 21-minute conference talk by Cocoa Xu at Code BEAM Europe 2023. Discover how TFLiteBEAM leverages TensorFlow Lite to create a compact, efficient solution for resource-constrained systems like Raspberry Pi. Learn about its small footprint of around 10MB, optimized performance for mobile and embedded devices, and compatibility with Edge TPU hardware accelerators such as Coral. Understand how TFLiteBEAM can benefit all BEAM languages with support for Erlang NIF libraries, offering a practical alternative to desktop GPUs for edge computing applications.