Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep neural network-based detection of counterfeit products from smartphone images

Hugo Garcia-Cotte, Dorra Mellouli, Abdul Rehman, Li Wang, David G. Stork

TL;DR

The world's first purely computer-vision-based system to combat counterfeiting-one that does not require special security tags or other alterations to the products or modifications to supply chain tracking is presented.

Abstract

Counterfeit products such as drugs and vaccines as well as luxury items such as high-fashion handbags, watches, jewelry, garments, and cosmetics, represent significant direct losses of revenue to legitimate manufacturers and vendors, as well as indirect costs to societies at large. We present the world's first purely computer-vision-based system to combat such counterfeiting-one that does not require special security tags or other alterations to the products or modifications to supply chain tracking. Our deep neural network system shows high accuracy on branded garments from our first manufacturer tested (99.71% after 3.06% rejections) using images captured under natural, weakly controlled conditions, such as in retail stores, customs checkpoints, warehouses, and outdoors. Our system, suitably transfer trained on a small number of fake and genuine articles, should find application in additional product categories as well, for example fashion accessories, perfume boxes, medicines, and more.

Deep neural network-based detection of counterfeit products from smartphone images

TL;DR

The world's first purely computer-vision-based system to combat counterfeiting-one that does not require special security tags or other alterations to the products or modifications to supply chain tracking is presented.

Abstract

Counterfeit products such as drugs and vaccines as well as luxury items such as high-fashion handbags, watches, jewelry, garments, and cosmetics, represent significant direct losses of revenue to legitimate manufacturers and vendors, as well as indirect costs to societies at large. We present the world's first purely computer-vision-based system to combat such counterfeiting-one that does not require special security tags or other alterations to the products or modifications to supply chain tracking. Our deep neural network system shows high accuracy on branded garments from our first manufacturer tested (99.71% after 3.06% rejections) using images captured under natural, weakly controlled conditions, such as in retail stores, customs checkpoints, warehouses, and outdoors. Our system, suitably transfer trained on a small number of fake and genuine articles, should find application in additional product categories as well, for example fashion accessories, perfume boxes, medicines, and more.
Paper Structure (8 sections, 4 figures, 1 table)

This paper contains 8 sections, 4 figures, 1 table.

Table of Contents

  1. The problem of counterfeit products
  2. Application goals and system specifications
  3. Recognition system, image data, and training and testing protocols
  4. Results
  5. Conclusion
  6. Response length
  7. Formatting your response
  8. Illustrations, graphs, and photographs

Figures (4)

  • Figure 1: Flowchart of our anti-counterfeiting system. A base image-classification deep network is transfer trained with a relatively small ($\approx 20,000$) examples of ground truth images of genuine and counterfeit product marks, preprocessed for affine transforms such as rotation, translations, and scale. The trained net is used to classify a query item as genuine or counterfeit.
  • Figure 2: (T) Images of genuine polo shirt emblems and (B) of counterfeit emblems.
  • Figure 3: The empirical tradeoff between rejection rate and classification accuracy for our AntiCounterfeit deep network classifier.
  • Figure 4: Example of caption. It is set in Roman so that mathematics (always set in Roman: $B \sin A = A \sin B$) may be included without an ugly clash.