V3I6P42

Admaximize: A Footfall-Driven Advertising Strategy

Riya H. S Pandey^1*, Shreeram G. Parab², Sandhya Kaprawan³

Abstract

Admaximize is a footfall-responsive advertising platform engineered to deliver context-aware advertisements by leveraging real-time and predicted crowd density at physical venues — retail stores, transit hubs, and commercial zones. The system integrates three components: (1) a YOLOv8-powered computer vision module for live people detection, (2) an ARIMA time- series forecasting component for near-term footfall estimation, and (3) a FastAPI-based microservices ingestion layer for continuous multi-site data processing. A structured review of 25 peer-reviewed papers, organised into five thematic clusters— object detection, time-series forecasting, data ingestion, crowd flow modelling, and multi-camera tracking — reveals ten significant research gaps. These gaps span: adaptive detection under dynamic conditions, event-driven forecasting failures, brittle ingestion pipelines, the wholly absent advertising decision layer, privacy compliance shortfalls, cross-camera identity continuity, cold-start forecasting, edge-versus-cloud benchmarking, single-modality estimation, and absence of a closed-loop feedback mechanism. This paper provides a structured gap analysis, traces each gap’s origin in the literature, and proposes targeted research directions for resolution.

Keywords:

Footfall analytics, YOLOv8, ARIMA, microservices, people counting, out-of-home advertising, crowd flow prediction, multi-camera tracking, privacy compliance, closed-loop systems