Kyocera Document Solutions Inc.

Kilho Shin

Dongpei Su

Sheng Li

Waqas Rasheed

Kendrick Esperanza Wong

A digital image comparator includes an Faster R-CNN configured to generate a filtered set of local feature maps of an input image, a match head, and logic to preserve the local feature maps in an indexed data structure and to make the local feature maps retrievable by the match head via local feature map indexes.

Computationally efficient document stamp detector

A method includes generating multiple attention map for an image from outputs of a corresponding different one of multiple convolutional hidden layers of a neural network. A different weighted attention map is then generated from each of the attention map. The weighted attention maps are input to a first fully-connected neural network layer to generate a colorfulness metric, which may be used to augment human-perceived colorfulness of the image.

Measurement and application of image colorfulness using deep learning

Embodiments provide an automated approach for generating unbiased synthesized image-label pairs for colorization training of retro photographs. Modern grayscale images with corresponding color images are translated to images with the characteristics of retro photographs, thereby producing training data that pairs images with the characteristics of retro paragraphs with corresponding color images. This training data can then be employed to train a deep learning model to colorize retro photographs more effectively.

Retro-to-modern grayscale image translation for preprocessing and data preparation of colorization

A digital image processor includes a region proposal network configured to transform digital image inputs into region proposals and bounding box refinement logic configured to transform the region proposals by determining a first set of the region proposals exhibiting dense spacing, determining a second set of the region proposals exhibiting sparse spacing, executing a first transformation to merge at least some of the region proposals exhibiting dense spacing to generate refined region proposals, executing a second transformation to join at least some of the region proposals exhibiting sparse spacing to generate additional ones of the refined region proposals, and applying an expansion transformation to the refined region proposals.

Consolidation of bounding boxes generated from text type pattern localizer

Methods and system are disclosed for a double-feed detector with neural network classifier. These includes accepting pages of a print-based media substrate into a scanning device with a paper feed device, back-illuminating the pages using a light source, capturing a page image of back-illuminated pages using an image capture device, detecting noise patterns exhibited by the page image using an image processor, classifying detected noise patterns using a noise pattern classifier, and determining, using classified detected noise patterns and a neural network controller for a neural network model, whether the at least one page is a sheet stack (at least two pages stacked together and passing through the paper feed device concurrently) based on the classified detected noise patterns. On condition a sheet stack is detected, a sheet stack detected signal triggers taking an action.

Method and apparatus for detecting sheet-fed scanner double-feeds using neural network classifier

A system employs a trained model to detect artifact(s) associated with artifact type(s) appearing in a reproduction of a source image (a test image). The system determines differences between the test image and the source image and outputs probabilities that the artifact(s) in the test image are associated with each of the artifact type(s). A dataset for training the model includes: (i) a reference category including reference image(s) without any artifacts; and (ii) artifact categories, each corresponding to a respective one of the artifact types and including noised images associated with the respective artifact type. Each noised image includes one of the reference images and an artifact associated with the respective artifact type. The model is trained to detect the artifact type(s) by providing the model with the dataset and causing the model to process differences between each noised image and the reference image in the noised image.

Growing community of inventors

Los Angeles, CA, United States of America

Kilho Shin