Abstract: A danger ranking training method comprising training a first deep neural network for generic object recognition within generic images, training a second deep neural network for specific object recognition within images of a specific application, training a third deep neural network for specific scene flow prediction within image sequences of the application, training a fourth deep neural network for potential danger areas localization within images or image sequences of the application using at least one human trained danger tagging method, training a fifth deep neural network for non-visible specific object anticipation and/or visible specific object prediction within image or image sequences of the application, and determining at least one danger pixel within an image or an image sequence of the application using an end-to-end deep neural network as a sequence of transfer learning of the five deep neural networks followed by one or several end-to-end top layers.

Abstract: Some embodiments are directed to a method to reinforce deep neural network learning capacity to classify rare cases, which includes the steps of training a first deep neural network used to classify generic cases of original data into specified labels; localizing discriminative class-specific features within the original data processed through the first deep neural network and mapping the discriminative class-specific features as spatial-probabilistic labels; training a second-deep neural network used to classify rare cases of the original data into the spatial-probabilistic labels; and training a combined deep neural network used to classify both generic and rare cases of the original data into primary combined specified and spatial-probabilistic labels.

Abstract: Some embodiments are directed to a processing method of a three-dimensional point cloud, including: obtaining a 3D point cloud from a predetermined view point of a depth sensor; extracting 3D coordinates and intensity data from each point of the 3D point cloud with respect to the view point; transforming 3D coordinates and intensity data into at least three two-dimensional spaces, namely an intensity 2D space function of the intensity data of each point, a height 2D space function of an elevation data of each point, and a distance 2D space function of a distance data between each point of 3D point cloud and the view point, defining a single multi-channel 2D space.

Abstract: A method for detecting contour points of an object in an image obtained by a video camera comprising the steps of: (i) selecting a scan line of the image; (ii) identifying minimum intensity differences called transitions between pixels of the selected scan line; (iii) identifying plateaus at both ends of the identified transitions; (iv) determining contour points of the object between the identified plateaus; (v) generating a descriptor of the contour in one dimension; and (vi) beginning again with step (i) by selecting another scan line of the image according to a predefined order.

Abstract: A communications receiver may include an adaptive filter unit for removing coherent interference components from a received signal. In the absence of a signal of interest, the filter may adapt dynamically to remove current interference components. When a signal of interest is detected, the filter may be controlled to stop (or at least reduce) its adaptation, to prevent removal of the signal of interest. The received signal may be down-converted to a complex baseband by conditioning circuitry. A detector may detect the signal of interest, and control the filter. Autocorrelation may be used to estimate a characteristic of the signal of interest in the complex baseband. The detector may include hysteresis to react quickly to the start of signal of interest, and more slowly to an end of the signal of interest. The signal of interest may be a frequency shift keyed (FSK) signal. A demodulator may demodulate FSK components based on the autocorrelation result.

Abstract: A communications receiver may include an adaptive filter unit for removing coherent interference components from a received signal. In the absence of a signal of interest, the filter may adapt dynamically to remove current interference components. When a signal of interest is detected, the filter may be controlled to stop (or at least reduce) its adaptation, to prevent removal of the signal of interest. The received signal may be down-converted to a complex baseband by conditioning circuitry. A detector may detect the signal of interest, and control the filter. Autocorrelation may be used to estimate a characteristic of the signal of interest in the complex baseband. The detector may include hysteresis to react quickly to the start of signal of interest, and more slowly to an end of the signal of interest. The signal of interest may be a frequency shift keyed (FSK) signal. A demodulator may demodulate FSK components based on the autocorrelation result.

Abstract: The present invention provides a compressor comprising a housing 1 defining an air inlet passage 2 and a volute duct. A rotary impeller 6 is located within the housing 1 between the inlet passage 2 and the volute duct. A sleeve 3 is mounted axially in the inlet passage 2 and divides the inlet passage into a radially outer portion 4 and a radially inner portion 5. A plurality of inlet guide vanes 8 are positioned in the radially outer portion 4 and impart a rotary component of movement pre-swirl) to air passing through the inlet passage 2. A fluid cut-off valve 9 is positioned in the radially inner portion 5 and for selectively preventing fluid flow therethrough and diverting all of the air through the radially outer portion 4 of the inlet passage 2 at low mass flow rates.

Abstract: The invention is related to a method for controlling a device for the ultrasound detection of objects in air, said device comprising a plurality of transducers, used as transmitters and adapted to operate in a phased grid, and at least one transducer, used as a receiver, which are mounted on a support, wherein, prior to at least one detection operation, a calibration step is carried out in order to ascertain a structural phase shift of the transmitter with respect to the reference transmitter, which is then memorized so as to later be able to correct the theoretical phase shifts of the transmitters from the structural phase shift value when shots are ordered later.