Advanced Smart City Monitoring: Real-Time Identification of Indian Citizen Attributes

TL;DR

The paper tackles real-time attribute recognition in Indian smart-city surveillance by evaluating three modeling approaches on a 600-image attribute dataset, addressing class imbalance and generalization. It compares BEiT, Swin Transformer, and a FeatClassifier (ResNet50 backbone with a custom head) and introduces ScaledBCELoss to mitigate imbalance, reporting that FeatClassifier offers the strongest generalization for real-time operation. A lightweight deployment pathway is demonstrated with a Tkinter GUI for live predictions and Hugging Face-based static predictions, signaling practical feasibility for city-scale monitoring. The work emphasizes privacy, ethical deployment, and potential benefits for safety and urban analytics, while outlining future directions in data, architecture, and multi-modal integration to enhance robustness and real-time performance.

Abstract

This project focuses on creating a smart surveillance system for Indian cities that can identify and analyze people's attributes in real time. Using advanced technologies like artificial intelligence and machine learning, the system can recognize attributes such as upper body color, what the person is wearing, accessories they are wearing, headgear, etc., and analyze behavior through cameras installed around the city.

Figures (4)

• Figure 1: Model Architecture and flow of the project
• Figure 2: Train & Validation Loss Vs Epoch model 1
• Figure 3: Train & Validation Loss Vs Epoch model 2
• Figure 4: Train & Validation Loss Vs Epoch model 3