Abstract: Devices, systems, and methods obtain data in a first modality; propagate the data in the first modality through a neural network, thereby generating network outputs, wherein the neural network includes a first-stage neural network and a second-stage neural network, wherein the first-stage neural network includes two or more layers, wherein each layer of the two or more layers of the first-stage neural network includes a plurality of respective nodes, wherein the second-stage neural network includes two or more layers, one of which is an input layer and one of which is an output layer, and wherein each node in each layer of the first-stage neural network is connected to the input layer of the second-stage neural network; calculate a gradient of a loss function based on the network outputs; backpropagate the gradient through the neural network; and update the neural network based on the backpropagation of the gradient.

Abstract: Devices, systems, and methods obtain a region of an image; generate known-object scores for the region using known-object detectors, wherein each known-object detector of the known-object detectors detects objects in a respective object class; determine a likelihood that the region includes a complete object; and determine a likelihood that the region includes an unknown object based on the likelihood that the region includes a complete object and on the known-object scores.

Abstract: Shape reconstruction of physical objects using structured light patterns and multiple layers of movable sheets which each include a transparent region and a pattern region. Shape reconstruction of specular regions on the surface of the object may proceed by a first phase in which a first sheet is moved to the pattern region and all other sheets are moved to the transparent region, and a first sequence of images is captured for patterns projected on the pattern region of the first sheet; and by a second phase in which the first sheet is moved to the transparent region and all other sheets are moved to the pattern region, and a second sequence of images is captured for patterns projected on the pattern region of the second sheet. Based on the captured images, the shape of the surface of an object, particularly objects having a mirror-like or specular surface having at least some specular characteristics, is reconstructed.

Abstract: Devices, systems, and methods for computer recognition of action in video obtain frame-level feature sets of visual features that were extracted from respective frames of a video, wherein the respective frame-level feature set of a frame includes the respective visual features that were extracted from the frame; generate first-level feature sets, wherein each first-level feature set is generated by pooling the visual features from two or more frame-level feature sets, and wherein each first-level feature set includes pooled features; and generate second-level feature sets, wherein each second-level feature set is generated by pooling the pooled features in two or more first-level feature sets, wherein each second-level feature set includes pooled features.

Abstract: Shape reconstruction of physical objects using structured light patterns and multiple layers of movable sheets which each include a transparent region and a pattern region. Shape reconstruction of specular regions on the surface of the object may proceed by a first phase in which a first sheet is moved to the pattern region and all other sheets are moved to the transparent region, and a first sequence of images is captured for patterns on the pattern region of the first sheet, such as by projection of patterns or pre-printed patterns; and by a second phase in which e the first sheet is moved to the transparent region and all other sheets are moved to the pattern region, and a second sequence of images is captured for patterns on the pattern region of the second sheet, such as by projection of patterns or pre-printed patterns.

Abstract: Devices, systems, and methods obtain data in a first modality; propagate the data in the first modality through a neural network, thereby generating network outputs, wherein the neural network includes a first-stage neural network and a second-stage neural network, wherein the first-stage neural network includes two or more layers, wherein each layer of the two or more layers of the first-stage neural network includes a plurality of respective nodes, wherein the second-stage neural network includes two or more layers, one of which is an input layer and one of which is an output layer, and wherein each node in each layer of the first-stage neural network is connected to the input layer of the second-stage neural network; calculate a gradient of a loss function based on the network outputs; backpropagate the gradient through the neural network; and update the neural network based on the backpropagation of the gradient.

Abstract: Devices, systems, and methods obtain a region of an image; generate known-object scores for the region using known-object detectors, wherein each known-object detector of the known-object detectors detects objects in a respective object class; determine a likelihood that the region includes a complete object; and determine a likelihood that the region includes an unknown object based on the likelihood that the region includes a complete object and on the known-object scores.

Abstract: Systems, devices, and methods for detecting light obtain a first light-measurement objective, wherein a light-measurement objective defines one or more of an intensity of emitted light, a spectrum of emitted light, a spectrum of detected light, and a reflectance angle of detected light; obtain a second light-measurement objective; obtain an angular-dependence objective, wherein the angular-dependence objective defines a difference between a first light-reflection angle and a second light-reflection angle; select first settings values for one or more first illumination-emission devices and one or more first illumination-detection devices based on the first light-measurement objective; select second settings values for one or more second illumination-emission devices and one or more second illumination-detection devices based on the second light-measurement objective; and select a speed of an object conveyor based on the angular-dependence objective.

Abstract: Devices, systems, and methods for computer recognition of action in video obtain frame-level feature sets of visual features that were extracted from respective frames of a video, wherein the respective frame-level feature set of a frame includes the respective visual features that were extracted from the frame; generate first-level feature sets, wherein each first-level feature set is generated by pooling the visual features from two or more frame-level feature sets, and wherein each first-level feature set includes pooled features; and generate second-level feature sets, wherein each second-level feature set is generated by pooling the pooled features in two or more first-level feature sets, wherein each second-level feature set includes pooled features.

Abstract: Systems, devices, and methods for detecting light obtain a first light-measurement objective, wherein a light-measurement objective defines one or more of an intensity of emitted light, a spectrum of emitted light, a spectrum of detected light, and a reflectance angle of detected light; obtain a second light-measurement objective; obtain an angular-dependence objective, wherein the angular-dependence objective defines a difference between a first light-reflection angle and a second light-reflection angle; select first settings values for one or more first illumination-emission devices and one or more first illumination-detection devices based on the first light-measurement objective; select second settings values for one or more second illumination-emission devices and one or more second illumination-detection devices based on the second light-measurement objective; and select a speed of an object conveyor based on the angular-dependence objective.

Abstract: Systems, devices, and methods for generating signatures for an image obtain an image, estimate a spectral image of the obtained image, calculate one or both of a detection component of the image and a residual component of the image, wherein the detection component of the image is based on the spectral image, a spectral-power distribution for a specific illuminant, and spectral sensitivities of a detector, and wherein calculating the residual component of the image is based on the spectral image, the spectral power distribution for the specific illuminant, and the spectral sensitivities of the detector or, alternatively, based on the spectral image and the calculated detection component, and generate an image signature based on one or both of the detection component and the residual component.

Abstract: The present disclosure relates to classification of a material type of an unknown material. The material type is classified into a probability distribution of multiple predetermined material types by using a collection of plural predetermined material classifiers. Each material type of the multiple predetermined material types is associated with a corresponding best performing classifier from the collection of plural predetermined material classifiers by a predesignated stored association. The plural material classifiers are applied to the unknown material to obtain a list of candidate material types. A list of potential best performing classifiers is looked up using the list of candidate material types as a reference into the predesignated stored association. A respective probability that the unknown material belongs to a material type is assigned based on the list of potential best performing classifiers.

Abstract: A property of a material is determined. The material is illuminated with a light beam of controlled spectral and coherence properties. A stack of speckle field images is recorded from speckle fields reflected from the illuminated material in multiple spectral channels. The stack of speckle field images includes multiple speckle field images each being recorded in a respectively different spectral channel. Statistical properties of the speckle field images in the stack of speckle field images are analyzed to determine at least one property of the illuminated material.

Abstract: Image capture using an image capture device that includes an imaging assembly having a spectral sensitivity tunable in accordance with a spectral capture mask, and imaging optics having a polarization filter tunable in accordance with a polarization control mask for projecting an image of a scene onto the imaging assembly. A default polarization control mask is applied to the imaging optics, and a default spectral capture mask is applied to the imaging assembly. An image of the scene is captured using the imaging assembly.

Abstract: The present disclosure relates to classification of a material type of an unknown material. The material type is classified into a probability distribution of multiple predetermined material types by using a collection of plural predetermined material classifiers. Each material type of the multiple predetermined material types is associated with a corresponding best performing classifier from the collection of plural predetermined material classifiers by a predesignated stored association. The plural material classifiers are applied to the unknown material to obtain a list of candidate material types. A list of potential best performing classifiers is looked up using the list of candidate material types as a reference into the predesignated stored association. A respective probability that the unknown material belongs to a material type is assigned based on the list of potential best performing classifiers.

Abstract: Systems and methods for simulating an image exposure capture a first set of images of a scene; and generate a flash preview image based on the first set of images, an estimated scene spectral reflectance, an estimated spectral power distribution of the scene, and one or more flash device settings, wherein the flash preview image approximates a brightness and a color appearance of a final image of the scene captured while the scene is illuminated by a flash device according to the one or more flash device settings.

Abstract: An image-capturing device and associated method for selective exposure adjustment of an image, wherein, image data of a scene is acquired, the scene is displayed on d display unit, the scene is divided into a plurality of regions, and a level of exposure is chosen for at least one region from the plurality of regions.

Abstract: Systems and methods for collaborative imaging include a device for collaborative image capturing device comprising a computational image sensor including a imaging sensor configured to detect light signals from a field of view, and one or more processors configured to control at least one parameter of the imaging sensor and to adjust the at least one imaging sensor parameter based on a respective light signal detected by one or more other computational image sensors, and a network interface configured to exchange data with the one or more other computational image sensors, wherein the exchanged data indicates the respective light signals detected by the one or more other computational image sensors.

Abstract: Compensation is performed for an image capture device which includes an image sensor which has a tunable spectral response and which is tunable in accordance with a capture mask. The compensation is for spatial non-uniformity in color sensitivity of the image sensor. A default capture mask is applied to the image sensor, and a sample image is captured using the image sensor tuned by the default capture mask. Color of the sample image is analyzed to identify spatial non-uniformity in color sensitivity of the image sensor. A compensation capture mask is constructed. The compensation capture mask is constructed using calculations based on the identified spatial non-uniformity so as to compensate for spatial non-uniformity in color sensitivity of the image sensor. The compensation capture mask is stored in a memory of the image capture device for application of the compensation capture mask to the image sensor.

Abstract: A method for adaptive spectral image capture that may be performed via an image capture device is disclosed. A default capture parameter is applied to an imaging assembly and a sample image of a scene is captured by an image capture device. The sample image is analyzed to identify transition zones between multiple different regions. An initial guess as to which spectral regions a filter mode might work best is obtained based on dominant transition region spectrum and a first iterated step in which numerical values for the filter mode are calculated. A second iterated step in which each such filter mode is evaluated for effectiveness against other filter modes. The regions in which a specific filter mode works best becomes associated with the filer mode and these regions become the guess for the next iteration.