AI_Hardware_Project_ECE6501_Group_4

Team Name:

(Enter your team name from Canvas)

• ECE6501-Group4

Team Members:

• Jiandi Wang
• Yunwei Cai
• Zilin Wang
• Henan Zhao

Project Title:

(Enter your project title - be creative)

• Accelerated Testing of Image Classification Task Based on Raspberry Pi 5

Project Description:

(Provide a short description of the problem you're addressing)

• The image classification task was accelerated on Raspberry Pi with Waveshare Hailo-8 accelerators.
• Comparing different models like MobileNet and ResNet, we show efficient AI reasoning on edge devices.

Key Objectives:

• Task Acceleration: Image classification task was accelerated on Raspberry Pi with Waveshare Hailo-8 accelerators.
• Model Comparison: Different models like MobileNet and ResNet were compared for efficient AI reasoning on edge devices.

Technology Stack:

(List the hardware platform, software tools, language(s), etc. you plan to use)

Hardware platform

• Raspberry Pi Start Guide
• M.2 Waveshare Hailo-8 Acce A; with PCIe To M.2 adaptor

Software Tools

• Raspberry Pi OS: The operating system running on the Raspberry Pi, providing a Linux-based environment for development.
• Hailo-8 SDK: Software development kit for Hailo-8, which includes libraries and tools for model conversion, optimization, and deployment on the Hailo-8 accelerator.
• TensorFlow Lite: Lightweight version of TensorFlow designed for mobile and embedded devices, used for running inference with optimized models.
• ONNX Runtime: A cross-platform, high-performance scoring engine for Open Neural Network Exchange (ONNX) models, enabling compatibility with various pre-trained models.
• Python: Main programming language for development, scripting, and model interaction.
• Visual Studio Code: For interactive testing, data analysis, and visualization of results.
• Matplotlib or Seaborn: Python libraries for data visualization, used to create graphs comparing inference time, FPS, accuracy, etc.
• psutil: Python library for monitoring CPU and memory usage during inference, used to gather resource usage statistics.

Languages

• Python: Used for writing scripts to load models, run inference, and collect performance metrics.

Data Sources

• Image Classification Datasets: Public datasets like CIFAR-10 or MNIST for model testing and benchmarking.

Expected Outcomes:

(Describe what you expect to deliver at the end of the project)

Compare the following metrics between different models:

• Inference time.
• Inference speed (frames per second).
• Accuracy, and CPU/memory utilization.
• The impact of quantization and optimization (if time permits).

Timeline:

(Provide a rough timeline or milestones for the project)

• About 5/12/2024