Computer Vision - Intelligent Video Analytics Application for Retail

Challenge

An innovative startup envisioned a cutting-edge application that could conduct intelligent video analytics for retail stores. The ambitions for this application included people counting, facial recognition, Automatic License Plate Recognition (ALPR), demographic analysis, queue time control, anomaly detection, and more. However, developing such a multifaceted application from scratch presented a massive challenge, especially when considering the requirements of compatibility with CPUS, GPUS, different cloud environments, and the creation of mobile and web apps for viewing dashboards, registers, and entries. Additionally, they sought to create a plug-in that could beautifully integrate with the software of major Video Management Systems (VMS) players to serve larger retail stores better.

Solution

candido.ai came on board to leverage their expertise in AI solutions to tackle the challenge head-on. A comprehensive team effort led to the successful creation of the envisioned application, incorporating all stipulated features efficiently. Users could now access real-time analytics data on queue times, demographic patterns, and even face and license plate recognition.

Not only was candido.ai able to design this application for local utilization, but it also ensured it ran flawlessly on multiple platforms, including CPUs, GPUs, and various Cloud environments—the end solution used exclusively open-source software that can be deployed on various platforms and cloud vendors. Inference time was a priority, so techniques like pruning and quantization were used, combined with high-performance inference software. Through dedicated mobile and web apps, users could easily visualize dashboards, register queues, entries, and more. Additionally, candido.ai developed a compatible plug-in to integrate seamlessly with major VMS player software, amplifying its reach and utility in large retail stores.

Outcome

The developed application underwent rigorous testing at one of the largest retailers in Brazil, with auspicious results. The intelligent video analytics application proved efficient, reliable, and capable of delivering valuable insights in real-time. Impressed by its capability, the retailer implemented it across all their stores, marking a milestone for this trailblazing startup. The successful implementation at such a large scale definitively validated the application’s efficacy, potential, and overall value proposition, laying a solid foundation for its wider adoption across the retail industry.