Self-Supervised Learning Across Domains.

Human adaptability relies crucially on learning and merging knowledge from both supervised and unsupervised tasks: the parents point out few important concepts, but then the children fill in the gaps on their own. This is particularly effective, because supervised learning can never be exhaustive and thus learning autonomously allows to discover invariances and regularities that help to generalize. In this paper we propose to apply a similar approach to the problem of object recognition across domains: our model learns the semantic labels in a supervised fashion, and broadens its understanding of the data by learning from self-supervised signals on the same images. This secondary task helps the network to focus on object shapes, learning concepts like spatial orientation and part correlation, while acting as a regularizer for the classification task over multiple visual domains. Extensive experiments confirm our intuition and show that our multi-task method, combining supervised and self-supervised knowledge, provides competitive results with respect to more complex domain generalization and adaptation solutions. It also proves its potential in the novel and challenging predictive and partial domain adaptation scenarios.

Environmental, field and impurity factors to increase the agricultural performance of Brazilian and Australian sugarcane mills.

This study aims to identify explanatory factors to increase the agricultural performance of Brazilian and Australian sugarcane mills. The relevance of Brazil and Australia for the sugar industry motivated the development this study based on the most important factors in both countries responsible for increasing the efficiency in sugarcane production. Thus, this study is designed to assess the hypothesis that there are a few explanatory variables that are deeply responsible for the agricultural efficiency in the sugar-energy sector. As a specific objective, it proposes a DEA (Data Envelopment Analysis) model that seeks to optimize the production of Total Recoverable Sugar (TRS) by planted area, and simultaneously, minimizes mineral and vegetable impurities. The sample consists of 82 observations from 32 sugarcane mills. An agricultural efficiency study was performed using the two-stage DEA, in which the evaluated mills according to the level of efficiency in the proposed model. Then, a Multiple Linear Regression Analysis was performed to identify the variables with the greatest influence on the performance of the mills in terms of efficiency. The results revealed six relevant variables for increasing the agricultural performance in the production of sugarcane: rainfall (mm weekly), chopped cane delivery (%), delivery time (h), borer (%), air humidity (%), and rods in raw wine (× 105/mL). Finally, semi-structured interviews with Brazilian and Australian experts in the sugar-energy sector allowed the identification of five other relevant complementary factors that were unavailable in the database: genetic variety, agricultural cultivation activities, edaphoclimatic factors, renewal of sugarcane fields and irrigation system. The results of this study were grouped into the dimensions of environment, yield, and impurities, providing quantification and better understanding of the identified explanatory factors and the agricultural performance in terms of production efficiency, offering fundamental information that enables managers to make decisions and prioritize the aspects that contribute more significantly to the increase in agricultural productivity of the planted area.

MultiDIAL: Domain Alignment Layers for (Multisource) Unsupervised Domain Adaptation.

One of the main challenges for developing visual recognition systems working in the wild is to devise computational models immune from the domain shift problem, i.e., accurate when test data are drawn from a (slightly) different data distribution than training samples. In the last decade, several research efforts have been devoted to devise algorithmic solutions for this issue. Recent attempts to mitigate domain shift have resulted into deep learning models for domain adaptation which learn domain-invariant representations by introducing appropriate loss terms, by casting the problem within an adversarial learning framework or by embedding into deep network specific domain normalization layers. This paper describes a novel approach for unsupervised domain adaptation. Similarly to previous works we propose to align the learned representations by embedding them into appropriate network feature normalization layers. Opposite to previous works, our Domain Alignment Layers are designed not only to match the source and target feature distributions but also to automatically learn the degree of feature alignment required at different levels of the deep network. Differently from most previous deep domain adaptation methods, our approach is able to operate in a multi-source setting. Thorough experiments on four publicly available benchmarks confirm the effectiveness of our approach.

Evaluation of the Italian transport infrastructures: A technical and economic efficiency analysis.

•The deregulation process played a key role in terms of airport efficiency.•The technical efficiency of 32 Italian airports has been evaluated•The methodological approach has been twofold: DEA and Tobit model•The main results show that efficiency is independent of an airport's size.•The findings highlight the decisive role of public shareholders in small airports.

Factors associated with urban non-fatal road-accident severity.

This paper reports on the factors associated with non-fatal urban-road accident severity. Data on accidents were gathered from the local traffic police in the City of Palermo, one of the six most populated cities in Italy. Findings from a mixed-effects logistic-regression model suggest that accident severity increases when two young drivers are involved, road traffic conditions are light/normal and when vehicles crash on a two-way road or carriageway. Speeding is more likely to cause slight or serious injury even when compared to a vehicle moving towards the opposite direction of traffic. An accident during the summer is more likely to result in a slight or serious injury than an accident during the winter, which is in line with evidence from Southern Europe and the Middle East. Finally, the severity of non-fatal accident injuries in an urban area of Southern Europe was significantly associated with speeding, the age of the driver and seasonality.

Poly(ester urethane) guides for peripheral nerve regeneration.

A biocompatible and elastomeric PU was synthesized from low-molecular-weight PCL as macrodiol, CMD as chain extender and HDI as chain linker for applications in the field of peripheral nerve repair. PU cast films supported in vitro attachment and proliferation of NOBEC. The in vitro adhesion and proliferation of S5Y5 neuroblastoma cells on the inner surface of uncoated, gelatin- and PL-coated PU guides were compared. Due to their superior in vitro performance, PL-coated PU guides were tested in vivo for the repair of 1.8 cm-long defects in rat sciatic nerves. The progressive regeneration was confirmed by EMG and histological analysis showing the presence of regenerating fibers in the distal stumps.

Melt-extruded guides for peripheral nerve regeneration. Part I: poly(epsilon-caprolactone).

Melt-extruded guides for peripheral nerve repair based on poly(epsilon-caprolactone) (PCL) were realised and their physico-chemical properties were evaluated. Preliminarily, PCL cast films were found to support the attachment and proliferation of Neonatal Olfactory Bulb Ensheating Cells (NOBEC). S5Y5 neuroblastoma cells were cultured inside PCL guides in their uncoated form or coated with a non-specific adhesion protein (gelatin) and a specific peptide for nerve regeneration (poly(L-lysine)). Coating increased cell density (gelatin) and/or the cell density rate on substrates (poly(L-lysine); gelatin) as compared to uncoated guides. Various in vivo tests were carried out for the repair of small (0.5 cm), medium (1.5 cm) and long (4.5 cm) size defects in the peripheral nerves of Wistar rats. For the small nerve defects, uncoated and coated PCL guides were tested. Results from in vivo tests were subjected to histological examination after 45 days, 6 and 8 months postoperative for small, medium and large defects, respectively. Regeneration was found for small and medium size defects. For 0.5 cm defects, the coating did not affect regeneration significantly. Grip-tests also evidenced functional recovery for the 1.5 cm-long defects treated with PCL guides, after 6 months from implantation. On the other hand, mechanical stiffness of PCL conduits impaired the repair of 4.5 cm-long defects in 8-month period: the lack of flexibility of the guide to rat movements caused its detachment from the implant site. The research showed that PCL guides can be used for the successful repair of small and medium size nerve defects, with possible improvements by suitable bio-mimetic coatings.

Ultrasound-guided retrieval of plant awns.

In dogs, surgical removal of plant awns causing a foreign body granuloma or abscess may be challenging. The inability to localize the foreign body during surgical removal often leads to abscess recurrence. In this report, we describe ultrasound-guided retrieval as an alternative to standard surgical retrieval in animals where the plant awn can be identified sonographically. This procedure was used in six dogs with a superficial abscess due to a plant awn, and in all dogs the foreign body was successfully retrieved with Hartmann forceps. No complication was observed and no abscess recurred. Minimally invasive ultrasound-guided plant awn retrieval appears to be a safe effective method of retrieving superficially located plant awns in the dog.

PAM-microfabricated polyurethane scaffolds: in vivo and in vitro preliminary studies.

Polymeric scaffolds were realised with linear degradable PU in the form of square, hexagonal and octagonal grids. They were characterised in terms of their mechanical properties. Analysis shows that the mechanical properties of the scaffolds depend on their geometries which are easily modulated using PAM. In vitro biological assays showed that PU promotes the adhesion and proliferation of fibroblast cells and that cell activities are better on PU scaffolds than on PU films. In vivo implantation of PU and PLGA scaffolds and PU films demonstrated that the scaffolds are completely resorbed after three months with a slight inflammatory response, while the PU film was still present after six months with an intense granulomatous reaction.

Suspension of bone marrow-derived undifferentiated mesenchymal stromal cells for repair of superficial digital flexor tendon in race horses.

It has been proven that mesenchymal stromal cells (MSCs) can differentiate into tenocytes. Attempts to repair tendon lesions have been performed, mainly using scaffold carriers in experimental settings. In this article, we describe the clinical use of undifferentiated MSCs in racehorses. Significant clinical recovery was achieved in 9 of 11 horses evaluated using ultrasound analysis and their ability to return to racing. Our results show that the suspension of a small number of undifferentiated MSCs may be sufficient to repair damaged tendons without the use of scaffold support. Ultrasound scanning showed that fibers were correctly oriented. By using undifferentiated cells, no ectopic bone deposition occurred. A sufficient number of cells was recovered for therapeutic purposes in all but 1 case. We suggest that the use of autologous MSCs is a safe therapeutic method for treating incompletely (i.e., not full-thickness) damaged tendons.