Abstract: The invention relates to a method (200) for training a dynamic-architecture convolutional neural network for semantic image segmentation, comprising a phase (102) of auto-annotation of images, to constitute a database (B1), called the first database, of auto-annotated images, comprising the following steps: applying (104) at least one transformation to at least one source image, and adding (106), to each transformed image thus obtained, a tag, called the first tag, corresponding to said transformation; said method (200) comprising at least one iteration of a phase (110) of training said neural network with the first image database (B1), and a second image database (B2) annotated with a tag, called a second tag, relating to the content of said image. It also relates to a computer program and a device implementing such a method, and a deep-learning dynamic-architecture convolutional neural network obtained by such a method.

Abstract: A process for the hetero-integration of a semiconductor material of interest on a silicon substrate, includes a step of structuring the substrate which comprises a step of producing a growth mask on the surface of the silicon substrate, the growth mask comprising a plurality of masking patterns, two masking patterns being separated by a trench wherein the silicon substrate is exposed; a step of forming a two-dimensional buffer layer made of a 2D material, the buffer layer being free of side bonds on its free surface and being formed selectively on at least one silicon plane of [111] orientation in at least one trench, the step of forming a buffer layer being performed after the structuring step; a step of forming at least one layer of a semiconductor material of interest on the buffer layer. The semiconductor material of interest is preferably a IV-IV, III-V, II-VI semiconductor material and/or a 2D semiconductor material.

Abstract: A device comprises at least one pair of excitation electrodes forming a capacitor; a floorplan (e.g., a ground plane); and an electronic circuit. The device comprises at least one control electrode arranged at a distance from the capacitor. A switching circuit, of the device, comprises a switch having an open state and a closed state. The switching circuit is designed to apply, to the control electrode, an electric potential common to the floorplan when the switch is in the closed state. The switching circuit is also designed to leave a floating electrical potential for the control electrode when the switch is in the open state. The electronic circuit is designed to measure the mutual capacitance between the pair of excitation electrodes when the switch is in the open state and when it is in the closed state.

Abstract: A method for determining a membership function, the membership functions allowing a stress level of a user to be determined on the basis of a physiological parameter measured on the user, the membership function varying, depending on the physiological parameter, between: a first value, representative of a rest state; and a second value, representative of a stressed state; the membership function taking into account a distribution function, which is defined beforehand, the distribution function being a continuous and monotonic function, the distribution function being applied to a normalized parameter established on the basis of a measured physiological parameter, the method comprising determining a normalization function, and optionally a standardization function, the normalization function and the optional standardization function being determined on the basis of physiological parameters measured on a plurality of test individuals, in various calibration periods.

Abstract: The present invention relates to determining the rheological parameters of fluids, and, more particularly, by means of a method using a continuous jet droplet generator.

Abstract: The present invention relates to a method of controlling a wind turbine by automatic online selection of a controller that minimizes the wind turbine fatigue. The method therefore relies on an (offline constructed) database (BDD) of simulations of a list (LIST) of controllers, and on an online machine learning step for determining the optimal controller in terms of wind turbine (EOL) fatigue. Thus, the method allows automatic selection of controllers online, based on a fatigue criterion, and switching between the controllers according to the measured evolution of wind condition.

Abstract: The invention relates to a method for purifying a carrier gas which comprises oxygenated impurities in a first oxidation degree, the purification method comprising the circulation, advantageously uninterrupted, of the carrier gas through and along a direction XX? of a filter, the filter being made of an oxygen scavenger material which has a redox potential E°, and of which a first portion is in a reduced redox state and within which the oxygenated impurities are scavenged and/or pass from the first oxidation degree to a second oxidation degree, the method further comprising the application to the filter of a greater electric potential ?V, as an absolute value, than the redox potential E° during a main purification cycle CP.

Abstract: A sensor may include a prism having a first face; a metal first layer covering, via a contact face, the first face; a light source; and a matrix-array detector; the device may include a dielectric second layer on which rests a transistor including a sheet made of a two-dimensional material, intended to form a channel region, a front face of the sheet comprising a specific functionalization via which specific targets are liable to be adsorbed, the specific functionalization being suitable for placing the adsorbed specific targets at a smaller distance Dd below which detection via electrical measurement by means of the specific transistor and via measurement of resonance of surface plasmons is possible.

Abstract: A filament type non-volatile memory device, includes a first electrode, a second electrode and an active layer extending between the first electrode and the second electrode, the active layer electrically interconnecting the first electrode to the second electrode, the device being suitable for having: a low resistive state, in which a conducting filament electrically interconnecting the first electrode to the second electrode uninterruptedly extends from end to end through the active layer, the filament having a low electric resistance, and a highly resistive state, in which the filament is broken, the filament having a high electric resistance. The device further includes a shunt resistance electrically connected in parallel to the active layer, between the first electrode and the second electrode.

Abstract: The invention relates to an industrial system comprising machines, systems for controlling machines connected by a first communication network, and a gateway intended to connect the first communication network to a second communication network. The gateway comprises a memory and comprises a processor configured to copy to the memory first data transmitted over the second communication network and relating to the operation of the machines.

Abstract: A light-emitting diode may include: a first n-doped semiconductor portion; a second p-doped semiconductor portion; an active zone disposed between the first and second portions and including at least one emitting semiconductor portion; a layer that is electrically conductive and optically transparent to at least one wavelength of the UV range configured to be emitted from the emitting portion, the layer being such that the second portion is disposed between the layer and the active zone. The semiconductors of the first portion and of the emitting portion may include compounds including nitrogen atoms as well as atoms of aluminum and/or of gallium. The semiconductor of the second portion may include AlX2Ga(1-X2-Y2)InY2N that is p-doped with magnesium atoms, wherein X2>0, Y2>0, and X2+Y2<1, and in which the atomic concentration of magnesium is greater than 1017 at/cm3. The electrically conductive layer may include doped diamond.

Abstract: A method for determining a region in which the actual concentration is located, in a medium, of a substrate made up of any molecule likely to undergo catalysed oxidation-reduction by a catalyst. The method includes the following steps: taking at least one group of at least two biosensors, each biosensor having a calibration curve of the signal induced by the oxidation-reduction reaction and having identical initial portions of their calibration curves up to a concentration value of the substrate from which the measurement of the signal differ; and when more than one group is present, the biosensors in different groups having different calibration curves without identical initial portions; placing the biosensors in contact with the medium; measuring the signal induced by the oxidation or reduction reaction for each biosensor in the group/groups; comparing all the signal values produced by the biosensors and following the method described in the description.

Abstract: The invention relates to a method for increasing the lifespan of animal sperm comprising contacting said sperm with an inhibitor of Slo3 potassium channel. The invention also relates to a use of an inhibitor of Slo3 potassium channel, for increasing the lifespan of animal sperm or motility of capacitated animal sperm, comprising contacting an inhibitor of Slo3 potassium channel with said sperm. Moreover, the invention relates to an artificial insemination instrument for use in artificial insemination of an animal, comprising animal sperm in contact with an inhibitor of Slo3 potassium channel. The invention also relates to a method for artificially inseminating an animal using said artificial insemination instrument. Eventually, the invention relates to a method for increasing the fertility of an animal, comprising contacting sperm of said animal with an inhibitor of Slo3 potassium channel; then artificially inseminating said animal with said sperm.

Abstract: A light-emitting diode is provided, including: a first layer of n-doped AlX1Ga(1-X1-Y1)InY1N, with X1>0 and X1+Y1?1; a second layer of p-doped AlX2Ga(1-X2-Y2)InY2N, with X2>0 and X2+Y2?1; an active area disposed between the first and the second layers and comprising at least one multi-quantum well emissive structure; nanowires based on AlN p-doped with indium and magnesium atoms, disposed on the second layer; and an ohmic contact layer in contact with the nanowires. A method for producing a light-emitting diode is also provided.

Abstract: The invention relates to a computer implemented method for real-time emotion recognition from a real-time audio signal. The method includes transcribing, into text, an audio speech signal contained in the audio signal by an automatic speech recognition model, and computing, by a speech representation model, a joint representation vector corresponding to a joint representation of the speech as a function of the speech signal and the text. The method also include computing, by an emotion prediction model, an emotion embedding vector as a function of the joint representation vector, and mapping the emotion in at least one emotional frame, according to the emotion embedding vector, by an emotion mapping model. The invention further relates to a computer program and a device implementing such a method.

Abstract: A device intended to be implanted in a human or animal body, in order to produce hydrogen in situ from molecules present in the body medium in which the device is implanted, this device having an anode and a cathode, which are each electrically connected to a pole of an electrical energy source, and having a semi-permeable material separating the electrodes from the body medium, in which device, when the connection to the electrical energy source is effective in situ, in the presence of body fluid, a closed electrical circuit is formed, with production of hydrogen at the cathode, the semi-permeable material having a cutoff threshold of between 50 and 500 Da.

Abstract: A method of generating training data for transferring knowledge from a trained artificial neural network to a further artificial neural network, the method including: a) injecting a first sample into the trained artificial neural network; b) reinjecting a pseudo sample, generated based on a replicated sample present at the one or more outputs of the trained artificial neural network, into the trained artificial neural network in order to generate a new replicated sample; and c) repeating b) one or more times, wherein the training data for training the further artificial neural network includes at least two of the reinjected pseudo samples originating from the same first sample and corresponding output values generated by the trained artificial neural network.

Abstract: The method for monitoring the electrical network comprises the identification of critical element(s) of the network, for at least one source-load pair that are connected to the network; the parametric estimation of a value of the electrical characteristic of each critical element, by using a numerical model modelling an electrical behavior of the network as a function of said electrical characteristic and of electrical measurements performed by means of sensors positioned in the network, a sensor being positioned in the network for each critical element on at least one network element or on at least one network path formed by a plurality of network elements affected by a disturbance in the electrical characteristic of this critical element; the evaluation, for said at least one source-load pair and the determination of a state of the electrical network by taking into account the operational criterion.

Abstract: The present invention relates to methods and pharmaceutical compositions for the treatment of diseases mediated by the NRP-1/OBR complex signaling pathway. In particular, the present invention relates to a method for treating a disease selected from the group consisting of cancers, obesity and obesity related diseases, anorexia, autoimmune diseases and infectious diseases in a subject in need thereof comprising administering the subject with a therapeutically effective amount of an antagonist of the NRP-1/OBR signaling pathway.

Abstract: A method for producing an aluminium nitride (AlN)-based layer on a structure with the basis of silicon (Si) or with the basis of a III-V material, may include several deposition cycles performed in a plasma reactor comprising a reaction chamber inside which is disposed a substrate having the structure. Each deposition cycle may include at least the following: deposition of aluminium-based species on an exposed surface of the structure, the deposition including at least one injection into the reaction chamber of an aluminium (Al)-based precursor; and nitridation of the exposed surface of the structure, the nitridation including at least one injection into the reaction chamber of a nitrogen (N)-based precursor and the formation in the reaction chamber of a nitrogen-based plasma. During the formation of the nitrogen-based plasma, a non-zero polarisation voltage Vbias_substrate may be applied to the substrate.