Abstract: This disclosure provides methods, devices, and systems for machine learning. The present implementations more specifically relate to automatons that can acquire input images and ground truth images for training neural network models. In some aspects, a system for acquiring training data may include a camera, an electronic display, and an apparatus configured to maintain the camera in a stationary position while moving the electronic display in and out of the camera's field-of-view (FOV). In some aspects, the system may further include a controller configured to acquire training data via the camera based on the positioning of the electronic display. In some implementations, the controller may acquire ground truth images of a scene while the electronic display is covering the camera's FOV. In some other implementations, the controller may acquire input images of the scene while the electronic display is outside the camera's FOV.

Abstract: This disclosure provides methods, devices, and systems for object detection. The present implementations more specifically relate to region of interest (ROI) inferencing techniques that can be implemented using a single object detection model. In some aspects, a computer vision system maps a set of grid cells to an input image so that each grid cell includes a respective portion of the image, and where each of the grid cells is assigned a respective priority value. The system selects an ROI of the image based on the priority value assigned to each grid cell and performs, on the ROI, an inferencing operation associated with an object detection model. The system updates the priority values for one or more of the grid cells based on a result of the inferencing operation. The system then selects another ROI based on the updated priority values and perform the inferencing operation on the new ROI.

Abstract: This disclosure provides methods, devices, and systems for low-light imaging. In some implementations, an image processor may be configured to reduce or remove noise associated with an image based, at least in part, on a neural network. For example, the neural network may be trained to infer a denoised representation of the image. In some aspects, the image processor may scale the brightness level of the image to fall within a normalized range of values associated with the neural network. In some other aspects, a machine learning system may scale the brightness levels of input images to match the brightness levels of ground truth images used to train the neural network. Still further, in some aspects, the machine learning system may scale the brightness levels of the input images and the brightness levels of the ground truth images to fall within the normalized range of values during training.

Abstract: This disclosure provides methods, devices, and systems for low-light imaging. In some implementations, an image processor may be configured to reduce or remove noise associated with an image based, at least in part, on a neural network. For example, the neural network may be trained to infer a denoised representation of the image. In some aspects, the image processor may scale the brightness level of the image to fall within a normalized range of values associated with the neural network. In some other aspects, a machine learning system may scale the brightness levels of input images to match the brightness levels of ground truth images used to train the neural network. Still further, in some aspects, the machine learning system may scale the brightness levels of the input images and the brightness levels of the ground truth images to fall within the normalized range of values during training.

Abstract: This disclosure provides methods, devices, and systems for low-light imaging. The present implementations more specifically relate to selecting images that can be used for training a neural network to infer denoised representations of images captured in low light conditions. In some aspects, a machine learning system may obtain a series of images of a given scene, where each of the images is associated with a different SNR (representing a unique combination of exposure and gain settings). The machine learning system may identify a number of saturated pixels in each image and classify each of the images as a saturated image or a non-saturated image based on the number of saturated pixels. The machine learning system may then select the non-saturated image with the highest SNR as the ground truth image, and the non-saturated images with lower SNRs as the input images, to be used for training the neural network.

Abstract: This disclosure provides methods, devices, and systems for training machine learning models. The present implementations more specifically relate to techniques for automating the annotation of data for training machine learning models. In some aspects, a machine learning system may receive a reference image depicting an object of interest with one or more annotations and also may receive one or more input images depicting the object of interest at various distances, angles, or locations but without annotations. The machine learning system maps a set of points in the reference image to a respective set of points in each input image so that the annotations from the reference image are projected onto the object of interest in each input image. The machine learning system may further train a machine learning model to produce inferences about the object of interest based on the annotated input images.

Abstract: A system and method for denoising a sequence of images while maintaining a consistent appearance among images displayed consecutively in the sequence. A machine learning system maps a first input image in the sequence of images to a first output image based on a neural network algorithm and determines a first network loss based on differences between the first output image and a ground truth image. The system further maps a second input image in the sequence of images to a second output image based on the neural network algorithm and determines a second network loss based on differences between the second output image and the ground truth image. The system determines a consistency loss based on differences between the first output image and the second output image and updates the neural network algorithm based on the first network loss, the second network loss, and the consistency loss.

Abstract: This disclosure provides methods, devices, and systems for low-light imaging. In some implementations, an image processor may be configured to reduce or remove noise associated with an image based, at least in part, on a neural network. For example, the neural network may be trained to infer a denoised representation of the image. In some aspects, the image processor may scale the brightness level of the image to fall within a normalized range of values associated with the neural network. In some other aspects, a machine learning system may scale the brightness levels of input images to match the brightness levels of ground truth images used to train the neural network. Still further, in some aspects, the machine learning system may scale the brightness levels of the input images and the brightness levels of the ground truth images to fall within the normalized range of values during training.

Abstract: This disclosure provides methods, devices, and systems for low-light imaging. The present implementations more specifically relate to selecting images that can be used for training a neural network to infer denoised representations of images captured in low light conditions. In some aspects, a machine learning system may obtain a series of images of a given scene, where each of the images is associated with a different SNR (representing a unique combination of exposure and gain settings). The machine learning system may identify a number of saturated pixels in each image and classify each of the images as a saturated image or a non-saturated image based on the number of saturated pixels. The machine learning system may then select the non-saturated image with the highest SNR as the ground truth image, and the non-saturated images with lower SNRs as the input images, to be used for training the neural network.

Abstract: Systems and methods for updating an enrollment template having a plurality of enrollment views of a biometric input object. A determination is made as to whether a new input biometric view is a candidate view for template update based on a match criterion, and a determination is made as to whether the new input biometric view increases coverage of the biometric input object by the enrollment template. The new input biometric view is added to the enrollment template as a new enrollment view in response to determining that the new biometric view i) is a candidate view for template update, and ii) increases coverage of the biometric input object by the enrollment template.

Abstract: Systems and methods for point-based image alignment are disclosed.

Abstract: Systems and methods for point-based image alignment are disclosed.

Abstract: Systems and methods for aligning images are disclosed.

Abstract: Systems and methods for aligning images are disclosed. A method includes receiving an input image of a biometric object; identifying a plurality of sets of candidate transformations, wherein each set of candidate transformations included in the plurality of sets of candidate transformations aligns the input image to a different enrollment image included in a plurality of enrollment images; grouping the plurality of sets of candidate transformations into a combined set of candidate transformations; selecting a subset of candidate transformations from the combined set of candidate transformations; and, identifying a first transformation based on the selected subset of candidate transformations, wherein the first transformation aligns the input image to a first enrollment image included in the plurality of enrollment images.

Abstract: Systems and methods for aligning images are disclosed.