SDK Components

The Processor SDK Linux for Edge AI can be divided into 3 parts, Applications, Processor SDK Linux and Processor SDK RTOS. Users can get the latest application updates and bug fixes from the public repositories (GitHub and git.ti.com) which aligns with the SDK releases done quarterly. One can also build every component from source by following the steps here, SDK Development flow

_images/edgeai-sdk-components.png

Fig. 20 Processor SDK Linux for Edge AI components

Edge AI Applications

The edge AI applications are designed for users to quickly evaluate various Deep Learning networks on TDA4 SoC either from TI Edge AI Model Zoo or a custom network by running standalone examples and Jupyter notebook applications. Once the network is finalized for performance and accuracy it can also be easily integrated in a typical capture-inference-display usecase using example GStreamer based applications for rapid prototyping and deployment.

edgeai-tidl-tools

This application repository provides standalone Python and C/C++ examples to quickly evaluate inference models using TFLite, ONNX and NeoAI-DLR runtime using file based inputs. It also houses the Jupyter notebooks similar to TI Edge AI Cloud which can be executed right on the TDA4 Starter Kit or EVM.

For more details on using this application repo please refer to the documentation and source code found here: https://github.com/TexasInstruments/edgeai-tidl-tools

edgeai-modelzoo

This repo provides collection of example Deep Neural Network (DNN) Models for various computer vision tasks. A few example models are packaged as part of the SDK to run out-of-box demos. More can be downloaded using a download script made available in the edge_ai_apps repo.

For more details on the the pre-imported models and related documentation please visit: https://github.com/TexasInstruments/edgeai-modelzoo

edge_ai_apps

These are plug-and-play Deep Learning applications which supports running open source runtime frameworks such as TFLite, ONNX and NeoAI-DLR with a live camera and display. They help connect realtime camera, video or RTSP sources to DL inference to live display, bitstream or RTSP sinks.

The latest source code with fixes can be pulled from: https://git.ti.com/cgit/edgeai/edge_ai_apps

edgeai-gst-plugins

This repo provides the source of custom GStreamer plugins which helps offload tasks to TDA4 hardware accelerators and advanced DSPs with the help of edgeai-tiovx-modules. The repo gets downloaded, built and installed as part of the Installing Dependencies step.

Source code and documentation: https://github.com/TexasInstruments/edgeai-gst-plugins

edgeai-tiovx-modules

This repo provides OpenVx modules which help access underlying hardware accelerators in the TDA4 SoC and serves as a bridge between GStreamer custom elements and underlying OpenVx custom kernels. The repo gets downloaded, built and installed as part of the Installing Dependencies step.

Source code and documentation: https://github.com/TexasInstruments/edgeai-tiovx-modules

Processor SDK RTOS

The Processor SDK Linux for Edge AI gets all the HWA drivers, optimized libraries, OpenVx framework and more from Processor SDK RTOS

For more information visit Processor SDK RTOS Getting Started Guide.

Processor SDK Linux

The Processor SDK Linux for Edge AI gets all the Linux kernel, filesystem, device-drivers and more from Processor SDK Linux

For more information visit Processor SDK Linux Software Developer’s Guide.