Neural Fashion Image Captioning : Accounting for Data Diversity

Image captioning has increasingly large domains of application, and fashion is not an exception. Having automatic item descriptions is of great interest for fashion web platforms hosting sometimes hundreds of thousands of images. This paper is one of the first tackling image captioning for fashion images. To contribute addressing dataset diversity issues, we introduced the InFashAIv1 dataset containing almost 16.000 African fashion item images with their titles, prices and general descriptions. We also used the well known DeepFashion dataset in addition to InFashAIv1. Captions are generated using the Show and Tell model made of CNN encoder and RNN Decoder. We showed that jointly training the model on both datasets improves captions quality for African style fashion images, suggesting a transfer learning from Western style data. The InFashAIv1 dataset is released on Github to encourage works with more diversity inclusion.

Local simulations of snow redistribution by wind with an intermediate-complexity snow cover model driven by different wind downscaling methods

<p>In mountainous terrain, wind-driven transport of deposited snow affects the overall distribution of snow, and can have a significant effect on snowmelt patterns even at coarser resolution.  In an operational modelling perspective, a compromise must be found to represent this complex small-scale process with enough accuracy while mitigating the computational costs of snow cover simulations over large domains. To achieve this compromise, we implemented the SNOWTRAN-3D snow transport module within the FSM intermediate complexity snow cover model. We included a new layering scheme and a historical variable of past snow wetting, but without resolving the snow microstructure. Simulations are run and evaluated over a small mountain range in the Swiss Alps at 25 to 100 m resolution. Being implemented in the model framework of the SLF operational snow hydrology service (OSHD), simulations further benefit from snow data assimilation techniques to provide improved estimates of solid precipitation fields. As complex wind patterns in mountains are the key processes driving snow transport, we tested statistical and dynamical methods to downscale 1 km resolution COSMO winds to better reflect topographically-induced flow patterns. These simulations are a first step working towards the integration of wind transport processes over large domains in an intermediate-complexity and -resolution operational modelling framework.</p>

How weakened cold pools open for convective self-aggregation

<p>In radiative-convective equilibrium (RCE) simulations, convective self-aggregation (CSA) is the spontaneous organization into segregated cloudy and cloud-free regions. Evidence exists for how CSA is stabilized, but how it arises favorably on large domains is not settled. Using large-eddy simulations (LES), we link the spatial organization emerging from the interaction of cold pools (CPs) to CSA. We systematically weaken simulated rain evaporation to reduce maximal CP radii, <em>R</em><sub>max</sub>, and find reducing <em>R</em><sub>max</sub> causes CSA to occur earlier. We further identify a typical rain cell generation time and a minimum radius, <em>R</em><sub>min</sub>, around a given rain cell, within which the formation of subsequent rain cells is suppressed. Incorporating <em>R</em><sub>min</sub> and <em>R</em><sub>max</sub>, we propose a toy model that captures how CSA arises earlier on large domains: when two CPs of radii <em>r</em><sub><em>i</em>,<em>j </em></sub>∈ [<em>R</em><sub>min</sub>, <em>R</em><sub>max</sub>] collide, they form a new convective event. These findings imply that CPs play a crucial role in RCE simulations by preventing the onset of CSA.</p>

Introduction to Bioentrepreneurship

Ever since the onset of civilization, humans have been continuously improvising and innovating on different aspects to make their livelihood more sustainable. This innovation reached a steep uptrend trajectory during the industrial revolution, and with the discovery of DNA and emergence of genetic engineering, our civilization entered a new era of biotechnology where desirable traits can be easily engineered into an organism to obtain a desired product. However, it was soon realized that in order to make the product reachable to the masses, integration of science, biology, and technology with business is of the utmost importance. This resulted in the genesis of the concept of bioentrepreneurship. Presently, it is a rapidly emerging discipline connecting two large domains, namely biotechnology and business. This chapter overviews bioentrepreneurship. It highlights the various requirements and measures implemented to start a bioentrepreneurship programme.

Editorial for Minerals Special Issue “From Diagenesis to Low-Grade metamorphism”

Rocks affected by pressure–temperature conditions in the transitional field between diagenesis and low-grade metamorphism make up large domains of the Earth’s upper continental and oceanic crust [...]

New Order-Revealing Encryption with Shorter Ciphertexts

As data outsourcing services have been becoming common recently, developing skills to search over encrypted data has received a lot of attention. Order-revealing encryption (OREnc) enables performing a range of queries on encrypted data through a publicly computable function that outputs the ordering information of the underlying plaintexts. In 2016, Lewi et al. proposed an OREnc scheme that is more secure than the existing practical (stateless and non-interactive) schemes by constructing an ideally-secure OREnc scheme for small domains and a “domain-extension” scheme for obtaining the final OREnc scheme for large domains. They encoded a large message into small message blocks of equal size to apply them to their small-domain scheme, thus their resulting OREnc scheme reveals the index of the first differing message block. In this work, we introduce a new ideally-secure OREnc scheme for small domains with shorter ciphertexts. We also present an alternative message-block encoding technique. Combining the proposed constructions with the domain-extension scheme of Lewi et al., we can obtain a new large-domain OREnc scheme with shorter ciphertexts or with different leakage information, but longer ciphertexts.