This paper examines the application of a deep learning approach to automatic coin recognition, via a mobile device and client-server architecture. We show that a convolutional neural network is effective for coin identification. During the training phase, we determine the optimum size of the training dataset necessary to achieve high classification accuracy with low variance. In addition, we propose a client-server architecture that enables a user to identify coins by photographing it with a smartphone. The image provided by the user is matched with the neural network on a remote server. A high correlation suggests that the image is a match. The application is a first step towards the automatic identification of coins and may help coin experts in their study of coins and reduce the associated expense of numismatic applications.
Implementation of a Coin Recognition System for Mobile Devices with Deep Learning
Capece, Nicola;Erra, Ugo;
2016-01-01
Abstract
This paper examines the application of a deep learning approach to automatic coin recognition, via a mobile device and client-server architecture. We show that a convolutional neural network is effective for coin identification. During the training phase, we determine the optimum size of the training dataset necessary to achieve high classification accuracy with low variance. In addition, we propose a client-server architecture that enables a user to identify coins by photographing it with a smartphone. The image provided by the user is matched with the neural network on a remote server. A high correlation suggests that the image is a match. The application is a first step towards the automatic identification of coins and may help coin experts in their study of coins and reduce the associated expense of numismatic applications.File | Dimensione | Formato | |
---|---|---|---|
sitis2016 CR.pdf
accesso aperto
Tipologia:
Documento in Pre-print
Licenza:
DRM non definito
Dimensione
8.96 MB
Formato
Adobe PDF
|
8.96 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.