Object Classification Research Prototype

This repository contains a research prototype for an object classification system. The system is designed to be deployed on both cloud and edge devices, and it uses a variety of models and technologies to achieve high performance and accuracy.

Table of Contents

• Architecture
• Getting Started
  • Prerequisites
  • Installation
  • Running the Services
• Deployment
• License
• References

Architecture

The architecture of the system is described in the architecture.drawio file. It consists of a set of microservices that work together to provide the object classification functionality.

Getting started

Prerequisites

• Docker
• Kubernetes (optional, for deployment)
• uv

Installation

1. Clone the repository:

   git clone <repository-url>
   cd object_classification_v2

2. Create a virtual environment and install dependencies:

   This project uses uv for package management. The dependencies are defined in each service's pyproject.toml file. To install the dependencies for all services, you can run the following command in the root directory:

   uv sync --all-packages

   Alternatively, you can install the dependencies for each service individually. For example, to install the dependencies for the preprocessing service:

   uv sync --package=preprocessing

Running the Services

The services can be run individually or all at once using the start_all_service.sh script.

To run all services:

./start_all_service.sh

To run individual services:

Each service has a run_server.sh script that can be used to start it. For example, to start the inference service:

cd src/inference
./run_server.sh

Deployment

The deployment directory contains scripts and configuration files for deploying the system to both cloud and edge environments. See the README files in those directories for more information.

License

This project is licensed under the terms of the Apache license.

References

If you use our prototype for research, please cite our paper where we describe the prototype in detail:

@INPROCEEDINGS{nguyen_2025_sagely,
  author={Nguyen, Hong-Tri and Yuan, Liang and Nguyen, Anh-Dung and Babar, M. Ali and Truong, Hong-Linh},
  booktitle={2025 IEEE International Conference on Web Services (ICWS)},
  title={SAGELY - Context-Aware Holistic Service Policy Enforcement Across Swarm-Edge Continuum},
  year={2025},
  pages={607-617},
  doi={10.1109/ICWS67624.2025.00083}}