Amazon Technologies, Inc.

Shahar Tsiper

Shai Mazor

Rahul Bhotika

Amit Adam

Oron Anschel

Hadar Averbuch Elor

Roee Litman

Ron Litman

Raghavan Manmatha

Stefano Soatto

Vijay Mahadevan

Jonathan Wu

Kunwar Yashraj Singh

Gal Sabina Star

Yair Kittenplon

Mohammed El Hamalawi

Niv Berchi Nayman

A document layout system for determining a layout of a document. The document layout system is configured to apply an OCR technique to identify the text in the document. The document layout system is further configured to generate a graph representation of the document, wherein the graph representation comprises a plurality of nodes and a plurality of edges that connect different ones of the plurality of nodes, wherein individual ones of the nodes correspond to different portions of the text. The document layout system is also configured to apply a graph cluster network machine learning model to the graph representation to identify a layout of different sections of the document according to respective merge inferences determined for individual ones of the plurality of edges. The document layout system is also configured to provide the layout of different sections of the document.

Graph-based document layout detection

Techniques for document rectification via homography recovery using machine learning are described. An image rectification system can intelligently make use of multiple pipelines for rectifying document images based on the detected type of device that generated the images. The image rectification system can provide high-quality rectifications without requiring human cooperation, multiple views of the document in multiple images, and/or without being constrained to only be able to process images from one source context.

Document rectification via homography recovery using machine learning

Techniques for recognizing text in an image are described. An exemplary method may include receiving a request to recognize text in an image; extracting features from the image and generating a visual feature sequence from the extracted features; performing selective contextual refinement at least one selective contextual refinement block of a stack of selective contextual refinement blocks to generate a text prediction by: generating a contextual feature map and combining the contextual feature map with the visual feature sequence into a visual feature space, and applying a selective decoder that utilizes a two-step attention on the visual feature space to generate a text prediction, wherein the two-step attention includes performing a 1-D self-attention computation to generate attentional features and decoding the attentional features to generate the text prediction; and outputting the generated text prediction.

Residual context refinement network architecture for optical character recognition

Techniques for auto-generation of annotated real-world training data are described. An electronic document is analyzed to determine text represented in the document and corresponding locations of the text. A representation of the electronic document is modified to include markers and printed. The printed document is photographed in real-world environments, and the markers within the digital photographs are analyzed to allow for the depiction of the document within the photographs to be rectified. The text and location data are used to annotate the rectified images.

Auto-annotation techniques for text localization

Growing community of inventors

Haifa, Israel

Shahar Tsiper