Patents by Inventor Paulo Mendonca
Paulo Mendonca has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).
-
Patent number: 11055819Abstract: Techniques for machine learning-based image super-resolution are described. A Dual Path Deep Back Projection Network can be used to enhance an input image. For example, the model may be trained to perform image super-resolution, remove artifacts, provide filtering or low light enhancement, etc. Classification may be performed on the resulting enhanced images to identify objects represented in the images. The model may be trained using a dataset that includes groups of images: an original image and an enhanced image. The model may use both residual and dense connectivity patterns between each successive back projection blocks to improve construction of a high-resolution output image from a low resolution input image. The enhanced images increase classification accuracy for input images having low image resolution.Type: GrantFiled: September 27, 2018Date of Patent: July 6, 2021Assignee: Amazon Technologies, Inc.Inventors: Kunwar Yashraj Singh, Paulo Mendonca
-
Patent number: 10796476Abstract: Techniques for improving a 2D to 3D image reconstruction network machine learning model are described. In some instances, this includes performing at least two transformations of a 3D model to generate at least two rotated 3D models, the at least two transformations to rotate the 3D model about an axis away from a viewing direction of the single 2D image; rendering the at least two rotated 3D models as rendered 2D images; and retraining a 2D to 3D image reconstruction network machine learning model using corresponding pairs of rotated 3D models and rendered 2D images.Type: GrantFiled: August 31, 2018Date of Patent: October 6, 2020Assignee: Amazon Technologies, Inc.Inventors: Yifan Xing, Paulo Mendonca, Rahul Tewari
-
Patent number: 10133951Abstract: Image data may be downloaded or otherwise obtained. An application executing on the computing device (or at least in communication with the computing device) can analyze the image data to recognize objects represented in the image data as well as determine locations or regions of the image that include the representations. Candidate regions of interest (e.g., bounding boxes) that include at least one detected object can be generated. A single region—a representative region—can be determined from the candidate regions based on user and/or application-defined criterion. The criteria may help ensure that the representative region includes statistical properties that at least satisfy threshold property levels, such as being a consistent estimator of the candidate regions. Notions of “average” or “median” regions can be applied to generate the representative region.Type: GrantFiled: October 27, 2016Date of Patent: November 20, 2018Assignee: A9.com, Inc.Inventors: Paulo Mendonca, Zhongwei Cheng
-
Publication number: 20170357858Abstract: A first map comprising local features and 3D locations of the local features is generated, the local features comprising visible features in a current image and a corresponding set of covisible features. A second map comprising prior features and 3D locations of the prior features may be determined, where each prior feature: was first imaged at a time prior to the first imaging of any of the local features, and lies within a threshold distance of at least one local feature. A first subset comprising previously imaged local features in the first map and a corresponding second subset of the prior features in the second map is determined by comparing the first and second maps, where each local feature in the first subset corresponds to a distinct prior feature in the second subset. A transformation mapping a subset of local features to a subset of prior features is determined.Type: ApplicationFiled: June 9, 2016Publication date: December 14, 2017Inventors: Paulo Mendonca, Christopher Brunner, Arvind Ramanandan, Murali Ramaswamy Chari, Dheeraj Ahuja
-
Patent number: 8311181Abstract: An imaging system includes an x-ray source configured to emit a beam of x-rays toward an object to be imaged, a detector configured to receive x-rays that are attenuated by the object, a data acquisition system (DAS) operably coupled to the detector, and a computer operably coupled to the DAS and programmed to obtain scan data with two or more incident energy spectra, decompose the obtained scan data into at least three basis materials, generate an image of one of the at least three basis materials using the decomposed scan data, and replace at least one pixel in the image using decomposed data of another of the at least three basis materials.Type: GrantFiled: February 1, 2010Date of Patent: November 13, 2012Assignee: General Electric CompanyInventors: Brian William Thomsen, Paulo Mendonca, Rahul Bhotika
-
Publication number: 20100135564Abstract: An imaging system includes an x-ray source configured to emit a beam of x-rays toward an object to be imaged, a detector configured to receive x-rays that are attenuated by the object, a data acquisition system (DAS) operably coupled to the detector, and a computer operably coupled to the DAS. The computer is programmed to obtain scan data with two or more incident energy spectra, identify one or more material triplet combinations based on data obtained prior to obtaining the scan data, decompose the obtained scan data into three or more basis materials based on the identified one or more material triplet combinations, and generate an image using the decomposed scan data.Type: ApplicationFiled: December 22, 2009Publication date: June 3, 2010Inventors: Brian William Thomsen, Paulo Mendonca, Rahul Bhotika
-
Publication number: 20100135565Abstract: An imaging system includes an x-ray source configured to emit a beam of x-rays toward an object to be imaged, a detector configured to receive x-rays that are attenuated by the object, a data acquisition system (DAS) operably coupled to the detector, and a computer operably coupled to the DAS and programmed to obtain scan data with two or more incident energy spectra, decompose the obtained scan data into at least three basis materials, generate an image of one of the at least three basis materials using the decomposed scan data, and replace at least one pixel in the image using decomposed data of another of the at least three basis materials.Type: ApplicationFiled: February 1, 2010Publication date: June 3, 2010Inventors: Brian William Thomsen, Paulo Mendonca, Rahul Bhotika
-
Publication number: 20060244830Abstract: Disclosed herein is a method that relates to mobile platform navigation through landmark recognition. The method including capturing images with at least one imaging device on the mobile platform and recognizing the captured images by comparing the captured images to images of landmarks stored in a database of location labeled landmark images. The method further including, navigating the mobile platform through tracking a location of the mobile platform relative to locations of the location labeled landmark images.Type: ApplicationFiled: June 30, 2006Publication date: November 2, 2006Inventors: David Davenport, Rahul Bhotika, Paulo Mendonca, Glenn Shaffer
-
Publication number: 20060215031Abstract: A method for calibrating a projective camera is provided. The method includes acquiring information by detecting at least one object on a substantially flat ground plane within a field of view. A projective camera calibration is performed. A measurement uncertainty is considered to yield a plurality of camera parameters from the projective camera calibration.Type: ApplicationFiled: December 7, 2005Publication date: September 28, 2006Inventors: Nils Krahnstoever, Paulo Mendonca, Jens Rittscher, Thomas Sebastian, Peter Tu
-
Publication number: 20060079743Abstract: A method includes generating a first response including a plurality of potential first objects of interest of a first shape from data regarding a region of interest (ROI), generating a second response including a plurality of second objects of interest for a second shape different than the first shape from the data regarding the ROI, and performing a morphological closing on the second response to produce a second cleaned response including a plurality of completed second objects. The method also includes decreasing the number of potential objects of interest in the first response by deleting any potential first objects that overlap any completed second objects to obtain a cleaned first response, and generating an image using at least one of the first cleaned response and the second cleaned response.Type: ApplicationFiled: October 8, 2004Publication date: April 13, 2006Inventors: Matthieu Ferrant, Saad Sirohey, Paulo Mendonca, Jerome Knoplioch, Rahul Bhotika
-
Publication number: 20060066911Abstract: An imaging system for correcting a bias in the location edges in an image is provided. The imaging system comprises an image processor configured to detect edges in an image of a given substructure, characterize a blurring factor in the image and correct a bias in the detected edges in the of a given substructure using the blurring factor.Type: ApplicationFiled: September 27, 2004Publication date: March 30, 2006Inventors: James Miller, Paulo Mendonca, Matthew Turek, Dirk Padfield
-
Publication number: 20050244042Abstract: The method comprises: accessing the multi-dimensional dataset; generating a plurality of differential operators for the multi-dimensional dataset using a discrete approximation of an analytic function; and forming a plurality of geometric responses based on a plurality of differential operators resultant from the generating. The method optionally further includes isolating a selected region of interest from the multi-dimensional dataset; the selected region of interest comprising a subset of the imaging volume.Type: ApplicationFiled: April 29, 2004Publication date: November 3, 2005Applicant: GENERAL ELECTRIC COMPANYInventors: Saad Sirohey, Rakesh Lal, Matthieu Ferrant, Paulo Mendonca
-
Publication number: 20050074162Abstract: A method of image reconstruction comprising: obtaining a Helmholtz reciprocal pair of images of an object, the images comprising a first image and a corresponding reciprocal image; determining an imaging geometry associated with the obtaining; selecting a plurality of points in the first image and identifying corresponding candidate points in the corresponding reciprocal image; matching a selected point of the plurality of points and a candidate point of the corresponding candidate points. A method of image registration with an object comprising: obtaining a Helmholtz reciprocal pair of images of an object, the Helmholtz reciprocal pair of images comprising a first image and a corresponding reciprocal image; estimating a pose for the object; predicting an estimated image corresponding to the pose and one image of the reciprocal pair of images; comparing the estimated image with a corresponding actual image from the pair of images; and refining the estimating a pose based on the comparing.Type: ApplicationFiled: October 3, 2003Publication date: April 7, 2005Inventors: Peter Tu, James Miller, Paulo Mendonca, James Ross