Abstract: A method and apparatus for performing deep neural network compression of convolutional and fully connected layers using a linear approximation of their outputs with information, such as in matrices representing weights, biases and non-linearities, to iteratively compress a pre-trained deep neural network by low displacement rank based approximation of the network layer weight matrices. Extension of the technique enables consecutive layers to be compressed jointly, allowing compression and speeding inference by reducing the number of channels/hidden neurons in the network.

Abstract: Techniques are disclosed for decreasing the time required for a photo-printing device (such as a commercial photo-printing kiosk) to print a plurality of digital images. For example, the digital images may be transferred from a print client to a print server and stored in a RAMdisk at the print server prior to printing. Image processing may be performed on at least some of the digital images prior to the initiation of printing. The digital images may be processed in decreasing order of estimated image processing time. Two or more of these techniques may be combined to decrease the total time required to print the digital images by, for example, increasing the speed with which images may be accessed, decreasing the amount of time that passes before printing begins, and enabling all of the digital images to be printed without stopping and restarting the print engine.

Abstract: A system and method for monitoring performance of an electromechanical system include monitoring a plurality of predetermined system conditions, collecting monitored values of the plurality of predetermined system conditions over time, and calculating system maintenance needs from collected monitored values.

Abstract: To avoid generating visible artifacts in the image, such as those generated when mildly defective areas are not identified, a system and method for identifying and correcting defects in a digital image including adjusting the pixel values of pixels surrounding the defective pixels are disclosed. The method for correcting defects in a input digital image comprises the steps of identifying the defects to form at least one defect map, generating a region of interest for each defect map, correcting the values of the pixels in each defect map, and adjusting the values of the pixels in each region of interest.

Abstract: A user of a system to correct defects in a digital image is provided with the ability to select, add, or verify defect areas through user input in order to prevent false detection of defects and complement the detection of defects obtained by analyzing the image. The user input includes identifying points as defects, whether as single defects or as a line of points (referred to as a scratch defect), defining areas of the input image where the defect identification is restricted to that area or precluded from that area, deleting defects from the at least one defect map, including additional defects in the at least one defect map by identifying at least one points as a defect, and selecting an area from the defect map display image and displaying the input image data in that area.