Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for estimating a 3-D pose of an object of interest from image and point cloud data. In one aspect, a method includes obtaining an image of an environment; obtaining a point cloud of a three-dimensional region of the environment; generating a fused representation of the image and the point cloud; and processing the fused representation using a pose estimation neural network and in accordance with current values of a plurality of pose estimation network parameters to generate a pose estimation network output that specifies, for each of multiple keypoints, a respective estimated position in the three-dimensional region of the environment.

Abstract: In one example embodiment, a computer-implemented method for extracting information from imagery includes obtaining data representing a sequence of images, at least one of the sequence of images depicting an object. The method includes inputting the sequence of images into a machine-learned information extraction model that is trained to extract location information from the sequence of images. The method includes obtaining as an output of the information extraction model in response to inputting the sequence of images, data representing a real-world location associated with the object depicted in the sequence of images.

Abstract: In one example embodiment, a computer-implemented method for extracting information from imagery includes obtaining data representing a sequence of images, at least one of the sequence of images depicting an object. The method includes inputting the sequence of images into a machine-learned information extraction model that is trained to extract location information from the sequence of images. The method includes obtaining as an output of the information extraction model in response to inputting the sequence of images, data representing a real-world location associated with the object depicted in the sequence of images.

Abstract: In one example embodiment, a computer-implemented method for extracting information from imagery includes obtaining data representing a sequence of images, at least one of the sequence of images depicting an object. The method includes inputting the sequence of images into a machine-learned information extraction model that is trained to extract location information from the sequence of images. The method includes obtaining as an output of the information extraction model in response to inputting the sequence of images, data representing a real-world location associated with the object depicted in the sequence of images.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for estimating a 3-D pose of an object of interest from image and point cloud data. In one aspect, a method includes obtaining an image of an environment; obtaining a point cloud of a three-dimensional region of the environment; generating a fused representation of the image and the point cloud; and processing the fused representation using a pose estimation neural network and in accordance with current values of a plurality of pose estimation network parameters to generate a pose estimation network output that specifies, for each of multiple keypoints, a respective estimated position in the three-dimensional region of the environment.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for determining estimated ground truth object keypoint labels for sensor readings of objects. In one aspect, a method comprises obtaining a plurality of sets of label data for a sensor reading of an object; obtaining respective quality control data corresponding to each of the plurality of sets of label data, the respective quality control data comprising: data indicating whether the labeled location of the first object keypoint in the corresponding set of label data is accurate; and determining an estimated ground truth location for the first object keypoint in the sensor data keypoint from (i) the labeled locations that were indicated as accurate by the corresponding quality control data and (ii) not from the labeled locations that were indicated as not accurate by the corresponding quality control data.

Abstract: Systems, methods, and computer readable media related to training and/or using a neural network model. The trained neural network model can be utilized to generate (e.g., over a hidden layer) a spectral image based on a regular image, and to generate output indicative of one or more features present in the generated spectral image (and present in the regular image since the spectral image is generated based on the regular image). As one example, a regular image may be applied as input to the trained neural network model, a spectral image generated over multiple layers of the trained neural network model based on the regular image, and output generated over a plurality of additional layers based on the spectral image. The generated output may be indicative of various features, depending on the training of the additional layers of the trained neural network model.

Abstract: A method including receiving an indication that a first classifier has identified that an image includes an object of a predetermined class of objects. Image data that relates to the image is processed using a second classifier with a first training state, which determines whether the image data includes the object of the predetermined class of objects. In response to the determining, data relating to the image data is transmitted to a remote system. Update data relating to the transmitted data is received from the remote system. The training state of the second classifier is updated to a second training state in response to the update data such that the second classifier with the second training state would make a different determination of whether future image data similar to the image data includes an object of the predetermined class of objects than the second classifier with the first training state.

Abstract: Systems, methods, and computer readable media related to training and/or using a neural network model. The trained neural network model can be utilized to generate (e.g., over a hidden layer) a spectral image based on a regular image, and to generate output indicative of one or more features present in the generated spectral image (and present in the regular image since the spectral image is generated based on the regular image). As one example, a regular image may be applied as input to the trained neural network model, a spectral image generated over multiple layers of the trained neural network model based on the regular image, and output generated over a plurality of additional layers based on the spectral image. The generated output may be indicative of various features, depending on the training of the additional layers of the trained neural network model.

Abstract: Systems, methods, and computer readable media related to training and/or using a neural network model. The trained neural network model can be utilized to generate (e.g., over a hidden layer) a spectral image based on a regular image, and to generate output indicative of one or more features present in the generated spectral image (and present in the regular image since the spectral image is generated based on the regular image). As one example, a regular image may be applied as input to the trained neural network model, a spectral image generated over multiple layers of the trained neural network model based on the regular image, and output generated over a plurality of additional layers based on the spectral image. The generated output may be indicative of various features, depending on the training of the additional layers of the trained neural network model.

Abstract: To process geological log data, a two-dimensional set of log values recorded at a plurality of points is obtained about a borehole periphery, and over a chosen length along the borehole. The set of log values are decomposed by identifying in the set one or more main functions indicative of one or more main geological layer features of the rock penetrated by the borehole and removing log values corresponding to values of the main functions from the set of log values. Further decomposing the set of log values can be performed, as necessary iteratively, based on one or more subsidiary functions. The log values of the set remaining after identification and removal of values are designated as texture. Functions indicative of the main and subsidiary layer features and the texture are used to construct a synthesized volume representing the rock removed during creation of the chosen length of the borehole.

Abstract: According to an aspect of the present disclosure, there is provided a method of image processing. The method comprises receiving image data comprising a set of feature vectors of a first dimensionality, the feature vectors corresponding to a class of objects. A variable projection is applied to each feature vector in the set of feature vectors to generate a set of projected vectors of a second dimensionality. The method then comprises processing the set of projected vectors to generate a model for the class of objects. A projection is applied to the model to generate an object classification model, of the first dimensionality, for the class of objects.

Abstract: Systems, methods, and computer readable media related to training and/or using a neural network model. The trained neural network model can be utilized to generate (e.g., over a hidden layer) a spectral image based on a regular image, and to generate output indicative of one or more features present in the generated spectral image (and present in the regular image since the spectral image is generated based on the regular image). As one example, a regular image may be applied as input to the trained neural network model, a spectral image generated over multiple layers of the trained neural network model based on the regular image, and output generated over a plurality of additional layers based on the spectral image. The generated output may be indicative of various features, depending on the training of the additional layers of the trained neural network model.

Abstract: A method including receiving an indication that a first classifier has identified that an image includes an object of a predetermined class of objects. Image data that relates to the image is processed using a second classifier with a first training state, which determines whether the image data includes the object of the predetermined class of objects. In response to the determining, data relating to the image data is transmitted to a remote system. Update data relating to the transmitted data is received from the remote system. The training state of the second classifier is updated to a second training state in response to the update data such that the second classifier with the second training state would make a different determination of whether future image data similar to the image data includes an object of the predetermined class of objects than the second classifier with the first training state.

Abstract: According to an aspect of the present disclosure, there is provided a method of image processing. The method comprises receiving image data comprising a set of feature vectors of a first dimensionality, the feature vectors corresponding to a class of objects. A variable projection is applied to each feature vector in the set of feature vectors to generate a set of projected vectors of a second dimensionality. The method then comprises processing the set of projected vectors to generate a model for the class of objects. A projection is applied to the model to generate an object classification model, of the first dimensionality, for the class of objects.

Abstract: To process geological log data, a two-dimensional set of log values recorded at a plurality of points is obtained about a borehole periphery, and over a chosen length along the borehole. The set of log values are decomposed by identifying in the set one or more main functions indicative of one or more main geological layer features of the rock penetrated by the borehole and removing log values corresponding to values of the main functions from the set of log values. Further decomposing the set of log values can be performed, as necessary iteratively, based on one or more subsidiary functions. The log values of the set remaining after identification and removal of values are designated as texture. Functions indicative of the main and subsidiary layer features and the texture are used to construct a synthesised volume representing the rock removed during creation of the chosen length of the borehole.

Abstract: The invention relates to a rotary screw machine of volume type comprising a body (30) having a main axis X, two members (10, 20), wherein a first one (20) surrounds a second one (10). Said first member (20) is hinged in said body (30) and is able to swivel on itself about its axis (Xf), aligned with said main axis X, according to a swiveling motion, whereas the axis (Xm) of said second member (10), revolves about the axis of said first member (Xf) according to an revolution motion having said length E as a radius. The machine further comprises a synchronizer (34, 36, 38, 40) synchronizing said swiveling motion and said revolution motion, such that a working medium performs a volumetric displacement in at least one working chamber (11) delimited by an outer surface (22) of said first member (20) and a inner surface (12) of said second member (10).

Abstract: In order to more effectively use the constructional volume of a volume screw machine of rotary type, a plurality of sets (80, 70; 60, 50) of female elements having an inner screw surface and of male elements having an outer screw surface is provided, wherein in each set a rotary motion of at least one element is created. If the motion of elements in different sets (80, 70; 60, 50) is synchronized, one can provide for a dynamically balanced machine.

Abstract: The Invention relates to a process for transforming heat and work in a thermoelectric cycle, wherein charge carriers of an electronic gas are cyclically subjected to at least a first and second (7) heat source. Thereby, heat is exchanged between elements of the cycle representing adjacent sections (c-d, d-a) of a thermodynamic representation of the thermoelectric cycle. The process can be performed without thermal loss and without thermal (entropy) degradation of the second source, which provides a thermoelectric efficiency higher than that of Carnot cycles.

Abstract: A method of transforming energy in a rotary screw machine that comprises a first and a second set of conjugated male and female elements spaced apart from each other along a central axis and having inner/outer profiled surfaces. Upon rotary motion of the male and/or female elements, working chambers are formed between these elements. The working chambers perform an axial movement. The rotary motions of the different sets (1, 2, 3) are synchronized in such a manner that synchronous and inphase motion of the elements in the different sets is performed with different values of angular periods of oscillation of axial movement of the working chambers. Thereby, a working medium transported in these working chambers can be compressed or expanded.