arXiv subjects

Ignazio Gallo

Publications and source records attributed to Ignazio Gallo.

23 records · Page 2Linked to original sources

arXiv2018

Seeing Colors: Learning Semantic Text Encoding for Classification

The question we answer with this work is: can we convert a text document into an image to exploit best image classification models to classify documents? To answer this question we present a novel text classification method which converts a text document into an encoded image, using word embedding and capabilities of Convolutional Neural Networks (CNNs), successfully employed in image classification. We evaluate our approach by obtaining promising results on some well-known benchmark datasets for text classification. This work allows the application of many of the advanced CNN architectures developed for Computer Vision to Natural Language Processing. We test the proposed approach on a multi-modal dataset, proving that it is possible to use a single deep model to represent text and image in the same feature space.

cs.CV↗

arXiv2018

Git Loss for Deep Face Recognition

Convolutional Neural Networks (CNNs) have been widely used in computer vision tasks, such as face recognition and verification, and have achieved state-of-the-art results due to their ability to capture discriminative deep features. Conventionally, CNNs have been trained with softmax as supervision signal to penalize the classification loss. In order to further enhance the discriminative capability of deep features, we introduce a joint supervision signal, Git loss, which leverages on softmax and center loss functions. The aim of our loss function is to minimize the intra-class variations as well as maximize the inter-class distances. Such minimization and maximization of deep features are considered ideal for face recognition task. We perform experiments on two popular face recognition benchmarks datasets and show that our proposed loss function achieves maximum separability between deep face features of different identities and achieves state-of-the-art accuracy on two major face recognition benchmark datasets: Labeled Faces in the Wild (LFW) and YouTube Faces (YTF). However, it should be noted that the major objective of Git loss is to achieve maximum separability between deep features of divergent identities.

cs.CV↗

arXiv2018

Revisiting Cross Modal Retrieval

This paper proposes a cross-modal retrieval system that leverages on image and text encoding. Most multimodal architectures employ separate networks for each modality to capture the semantic relationship between them. However, in our work image-text encoding can achieve comparable results in terms of cross-modal retrieval without having to use a separate network for each modality. We show that text encodings can capture semantic relationships between multiple modalities. In our knowledge, this work is the first of its kind in terms of employing a single network and fused image-text embedding for cross-modal retrieval. We evaluate our approach on two famous multimodal datasets: MS-COCO and Flickr30K.

cs.CV↗

arXiv2017

e-Fair: Aggregation in e-Commerce for Exploiting Economies of Scale

In recent years, many new and interesting models of successful online business have been developed, including competitive models such as auctions, where the product price tends to rise, and group-buying, where users cooperate obtaining a dynamic price that tends to go down. We propose the e-fair as a business model for social commerce, where both sellers and buyers are grouped to maximize benefits. e-Fairs extend the group-buying model aggregating demand and supply for price optimization as well as consolidating shipments and optimize withdrawals for guaranteeing additional savings. e-Fairs work upon multiple dimensions: time to aggregate buyers, their geographical distribution, price/quantity curves provided by sellers, and location of withdrawal points. We provide an analytical model for time and spatial optimization and simulate realistic scenarios using both real purchase data from an Italian marketplace and simulated ones. Experimental results demonstrate the potentials offered by e-fairs and show benefits for all the involved actors.

cs.CE↗

arXiv2016

Fairs for e-commerce: the benefits of aggregating buyers and sellers

In recent years, many new and interesting models of successful online business have been developed. Many of these are based on the competition between users, such as online auctions, where the product price is not fixed and tends to rise. Other models, including group-buying, are based on cooperation between users, characterized by a dynamic price of the product that tends to go down. There is not yet a business model in which both sellers and buyers are grouped in order to negotiate on a specific product or service. The present study investigates a new extension of the group-buying model, called fair, which allows aggregation of demand and supply for price optimization, in a cooperative manner. Additionally, our system also aggregates products and destinations for shipping optimization. We introduced the following new relevant input parameters in order to implement a double-side aggregation: (a) price-quantity curves provided by the seller; (b) waiting time, that is, the longer buyers wait, the greater discount they get; (c) payment time, which determines if the buyer pays before, during or after receiving the product; (d) the distance between the place where products are available and the place of shipment, provided in advance by the buyer or dynamically suggested by the system. To analyze the proposed model we implemented a system prototype and a simulator that allow to study effects of changing some input parameters. We analyzed the dynamic price model in fairs having one single seller and a combination of selected sellers. The results are very encouraging and motivate further investigation on this topic.

econ.GN↗