Paddle Examples for Arm Virtual Hardware(AVH)

Arm Virtual Hardware(AVH)

Arm Virtual Hardware (AVH) scales and accelerates IoT software development by virtualising popular IoT development kits, Arm-based processors, and systems in the cloud. It is an evolution of Arm’s modelling technology that removes the wait for hardware and the complexity of building and configuring board farms for testing. It enables modern agile software development practices, such as DevOps and MLOps workflows.

Arm Virtual Hardware is available for Corstone platforms and Cortex processors via an Amazon Machine Image (AMI) on AWS Marketplace as well for third-party hardware available via Arm’s SaaS platform.

For examples in this repository, we use Arm Virtual Hardware with Corstone platforms and Cortex processors via AWS.

PaddlePaddle

PaddlePaddle (PArallel Distributed Deep LEarning) is a simple, efficient and extensible deep learning framework developed by Baidu, Inc. As the first independent R&D deep learning platform in China, it has been officially open-sourced to professional communities since 2016. It is an industrial platform with advanced technologies and rich features that cover core deep learning frameworks, basic model libraries, end-to-end development kits, tools & components as well as service platforms. For more details, please refer to PaddlePaddle Github for details.

Example

We provide 4 use cases in this reposiotry (ocr, object_classification, object_detection and object_segmentation) To run the demos in Arm Virtual Hardware context, please follow these 3 steps:

1. Set up running environment

When you try to run the demo for the first time, you need to set up the running environment in AVH instance by the following command.

cd /path/to/Paddle-examples-for-AVH
sudo bash scripts/config_cmsis_toolbox.sh
sudo bash scripts/config_tvm.sh

2. Run the demo

You must specify the model name (by parameter --model) and the device name (by parameter --device) when you run the demo.

cd /path/to/usecase
bash run_demo.sh --model model_name --device device_name

Parameter options can be found in the following table.

Use Case	Model Name	Device Name
object_classification	MobileNetV3 PP_LCNet MobileNetV1	cortex-m55 cortex-m85
ocr	CH_PPOCRV2_CLS EN_PPOCRV3_REC	cortex-m55 cortex-m85
object_detection	Picodet	cortex-m55 cortex-m85
object_segmentation	PP_HumanSeg	cortex-m55 cortex-m85

For example, to run object classfication demo with PP-LCNet model on Arm Cortex-M55 platform, input the following command:

bash run_demo.sh --model PP_LCNet --device cortex-m55