RetailAI Services

AI-Powered Retail Traffic Analytics

Footfall and dwell-time analytics from existing CCTV for a regional mall operator in Thailand.

Key Details

Challenge	A regional shopping-mall operator needed footfall and dwell-time analytics from the CCTV it already had, rather than a new sensor install.
Solution	Multi-stream analytics on existing CCTV: people detection and tracking, a traffic engine for footfall and dwell time, and a heatmap dashboard.
Technologies	YOLO, OpenCV, PyTorch, TensorFlow, Python, Node.js, React, TypeScript, FFmpeg, AWS EC2, S3, Lambda, Kafka, Redis

Technologies used

YOLOOpenCVPyTorchTensorFlowPythonNode.jsReactTypeScriptFFmpegAWSKafkaRedis

Client

The client is a regional shopping-mall operator in Thailand, running large retail and entertainment complexes across Southeast Asia. The work turns existing CCTV into footfall, dwell-time and queue analytics for layout and operations.

Product screenshots

Light analytics web app showing a high-angle mall corridor camera feed with a bounding box and ID on each shopper, short motion trails, a highlighted counting line across the walkway with in and out counters, translucent dwell zones over two store entrances, plus a side panel listing live occupancy, zone dwell times and the camera list. — Live camera view: shoppers are tracked with persistent IDs as they cross a counting line, with entry and exit counts and dwell time at each store front.

Dark floor-plan view of a shopping mall level with store outlines, a glowing footfall heatmap concentrated around the atrium and escalators, arrows showing the most common shopper paths, a time slider, and a side panel ranking zones by visits and average dwell time with a hourly comparison chart. — Floor analytics: tracks from every camera are merged onto the floor plan as a footfall heatmap, with the most common paths between stores.

What we built

Multi-stream analytics on existing CCTV, with no extra sensors
People detection, tracking and movement analysis
Traffic engine for customer flow, zone occupancy and dwell time in stores and corridors
Queue detection at checkouts and food courts
Heatmaps, traffic statistics, real-time alerts and historical analytics

Engagement

Delivery is time and materials with 6 engineers and is ongoing.

Project team: 6 engineers (1 AI, 1 computer vision, 2 backend, 1 frontend, 1 QA). Duration: Ongoing. Model: Time & materials.

Services used

AI Services →

Industries

Retail & E-commerce →