VARTEQ

RETAIL & ECOMMERCE CASE STUDY

AI product recognition case study

A US consumer goods company needed a clearer view of where its products appeared in video ads and retail displays. The team developed a computer vision API that detects products in images and video, connects the results to the client's analytics system and reduces manual review.

TRANSFORM WITH VARIOUS TECHNOLOGIES

At a glance

Project type
Product Recognition API
Client
Undisclosed consumer goods company
Industry
Retail & eCommerce
Location
United States
Services
AI Development, Custom Software Development
Product detection accuracy
97.2%
Video frames processed per second
100+
Reduction in manual auditing costs
87%

OVERVIEW

Project overview

The company needed to assess product visibility across two sources: advertising videos and images of retail shelves. Manual reviews took time, and the results were difficult to use for timely campaign and placement decisions.

The project introduced an image and video analysis service built around a Product Recognition API. CNN models identify products in the source material, and the API makes the results available to the client's business intelligence workflow.

WHERE THEY STARTED

The challenge

Manual review

A product might appear for only part of a video ad or occupy a small area of a store shelf. To evaluate those placements, the client relied on manual checks of images and footage. That work was slow, and inconsistent identification made it difficult to compare results across campaigns and retail locations.

Existing recognition tools

The team also assessed existing image-recognition options and found they did not meet the project's need for precise detection and timely product visibility data.

WHAT WE BUILT

The solution

  1. Product detection with computer vision

    CNN models trained to recognize the client's products in images and video frames. The team tested detection performance before integration with the analytics workflow.

  2. Product Recognition API

    An API built to process the image-analysis results and make product detection data available to the rest of the client's system. TensorFlow supported the computer vision work.

  3. GPU-backed media processing

    GPU-accelerated cloud infrastructure supported image and video analysis. The system processed video frames for product detection, without a manual review of each frame.

  4. Analytics integration

    The solution connected with the client's business intelligence system, which could use the detection results in product visibility reports and advertising reviews.

SERVICES ON THIS PROJECT

TECHNOLOGIES USED

  • TensorFlow
  • Convolutional neural networks (CNNs)
  • GPU-accelerated cloud infrastructure
  • Product Recognition API

WHAT CHANGED

The outcome

  • Product appearances in ad footage and retail images could be checked through an automated recognition workflow rather than manual review alone.
  • Detection results became available to the client's analytics system for advertising and shelf-placement analysis.
  • 97.2% product detection accuracy and more than 100 video frames processed per second.
  • 87% reduction in manual auditing costs after automation.

More client stories

All case studies

START THE CONVERSATION

Facing a similar challenge?

Tell us what your team needs to detect, measure or verify across images and video. A VARTEQ partner will reply to arrange a first call.