Person Re-Identification

1️⃣ Introduction

What is Person Re-Identification (Re-ID)?

Person Re-Identification (Re-ID) is a computer vision task that aims to recognize and match individuals across multiple non-overlapping camera views. It is widely used in surveillance, pedestrian tracking, and smart city applications. The main challenges in Re-ID include variations in lighting, pose, occlusion, and background clutter.

About this Project

This project focuses on implementing Person Re-ID using OSNet, a state-of-the-art lightweight deep learning model specifically designed for re-identification tasks. OSNet effectively captures multi-scale discriminative features while maintaining high efficiency.

Key Features

• Deep Learning Framework: PyTorch
• Backbone Model: OSNet
• Datasets Used: CUHK03
• Evaluation Metrics: Mean Average Precision (mAP), Rank-1 Accuracy
• Applications: Security surveillance, pedestrian analysis, forensic investigations

Example Use Case

Given a query image of a person, the model searches a gallery dataset to find the most similar identities, ranking them based on similarity scores.

Example visualization:

This repository provides an end-to-end pipeline for training, evaluating, and deploying a Re-ID model using OSNet. In the following sections, I will guide you through the setup, training, and evaluation process.

2️⃣ Installation & Setup

Prerequisites

Before running the project, ensure you have the following dependencies installed:

• Python
• PyTorch
• Torchvision
• NumPy
• OpenCV
• Matplotlib
• scikit-learn
• tqdm

You can install them using:

pip install -r requirements.txt

Setting Up the Dataset

1. Download the CUHK03 dataset from the official sources.
2. Extract the dataset and place it in the data/ folder.
3. Ensure the dataset folder structure follows:

   data/
   ├── cuhk03/
   │   ├── archive
   │   │    ├── ...
   │   ├── pairs.csv
   

Running the Code

To train the model, use:

python src/trainer.py

To evaluate the model, run:

updating...

To perform inference on a new image:

updating...