In this paper we investigate the problem of Grocery product recognition using iconic images. Iconic images are used to advertise products and they are very different from images that are captured in-store. We investigate the use of learned features for the retrieval process. We evaluated different feature extraction strategies using Convolutional Neural Networks (CNNs) and tested the CNNs on the Grocery Store image dataset that contains 81 product categories grouped into 43 coarse-grained classes and 3 macro classes. Results show that a Siamese network with a DenseNet-169 backbone better captures relations between iconic and in-store images outperforming other architectures in the retrieval task.
Ciocca, G., Napoletano, P., Locatelli, S. (2021). Iconic-Based Retrieval of Grocery Images via Siamese Neural Network. In Pattern Recognition. ICPR International Workshops and Challenges (pp.269-281). Springer Science and Business Media Deutschland GmbH [10.1007/978-3-030-68790-8_22].
Iconic-Based Retrieval of Grocery Images via Siamese Neural Network
Ciocca, Gianluigi
;Napoletano, Paolo;
2021
Abstract
In this paper we investigate the problem of Grocery product recognition using iconic images. Iconic images are used to advertise products and they are very different from images that are captured in-store. We investigate the use of learned features for the retrieval process. We evaluated different feature extraction strategies using Convolutional Neural Networks (CNNs) and tested the CNNs on the Grocery Store image dataset that contains 81 product categories grouped into 43 coarse-grained classes and 3 macro classes. Results show that a Siamese network with a DenseNet-169 backbone better captures relations between iconic and in-store images outperforming other architectures in the retrieval task.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
