tensorflow-lite-raspberrypi

Precompiled TensorFlow Lite 2.11.0 binaries for Raspberry Pi 3 & 4. Read the following blog post for additional information.

Supported features

• NEON optimization
• VFPv4 optimization
• XNNPACK delegate
• Ruy matrix multiplication library
• MMAP-based allocation
• C and C++ APIs
• Python 3 bindings

Prerequisites

Supported Boards

• Raspberry Pi 3 Model A+
• Raspberry Pi 3 Model B+
• Raspberry Pi 4 Model B

Tested on Raspberry Pi 4 Model B (8 GB).

Supported OS

• Raspberry Pi OS Bullseye (32-bit and 64-bit)

Install

• 32-bit:

wget https://github.com/prepkg/tensorflow-lite-raspberrypi/releases/latest/download/tensorflow-lite.deb

sudo apt install -y ./tensorflow-lite.deb

• 64-bit:

wget https://github.com/prepkg/tensorflow-lite-raspberrypi/releases/latest/download/tensorflow-lite_64.deb

sudo apt install -y ./tensorflow-lite_64.deb

Uninstall

sudo apt purge --autoremove -y tensorflow-lite

Debian Package

Debian package contains the following shared libraries:

Library	Description
libtensorflowlite_c.so	C API to access TensorFlow Lite interpreter and perform an inference
libtensorflow-lite.so	C++ API to access TensorFlow Lite interpreter and perform an inference

Reference

1. TensorFlow Lite repository