Tata Consultancy Services Limited

Monika Sharma

Lovekesh Vig

Shubham Paliwal

Rohit Rahul

Ramya Sugnana Murthy Hebbalaguppe

Hiranmay Ghosh

Arindam Chowdhury

Manu Sheoran

Gaurav Gupta

Gautam Shroff

Shirish Subhash Karande

Ehtesham Hassan

Ashwin Srinivasan

Anand Sriraman

Gunjan Sehgal

Vishwanath Doreswamy

Swati

Swati Jindal

Ankit Verma

Kiran Francis

Vishwanath Doreswamy Gowda

Arushi Jain

Meghal Dani

Shikha Gupta

Kushagra Mahajan

The present disclosure detects lesions in different datasets using a semi-supervised domain adaptation manner with very few labeled target samples. Conventional approaches suffer due to domain-gap between source and target domain. Initially, the system receives an input image, and extracts a plurality of multi-scale feature maps from the input image. Further, a classification map is generated based on the plurality of multi-scale feature maps. Further, a 4D vector corresponding to each of a plurality of foreground pixels is computed. Further, an objectness score corresponding the plurality of foreground pixels is computed. After computing the objectness score, a centerness score is computed for each of the plurality of foreground pixels using a single centerness network. Further, an updated objectness score is computed for each of the plurality of foreground. Finally, a plurality of multi-sized lesions in the input image are detected using a trained few-shot adversarial lesion detector network.

Method and system for semi-supervised domain adaptation based universal lesion detection

Keypoint extraction is done for extracting keypoints from images of documents. Based on different keypoint extraction approaches used by existing keypoint extraction mechanisms, number of keypoints extracted and related parameters vary. Disclosed herein is a method and system for keypoint extraction from images of one or more documents. In this method, a reference image and a test image of a document are collected as input. During the keypoint extraction, based on types of characters present in words extracted from the document images, a plurality of words are extracted. Further, all connected components in each of the extracted words are identified. Further, it is decided whether keypoints are to be searched in a first component or in a last component of all the identified connected components, and accordingly searches and extracts at least four of the keypoints from the test image and the corresponding four keypoints from the reference image.

Method and system for keypoint extraction from images of documents

State of the art deep network based Universal Lesion Detection (ULD) techniques inherently depend on large number of datasets for training the systems. Moreover, these system are specifically trained for lesion detection in organs of a Region of interest (RoI) of a body. Thus, requires retraining when the RoI varies. Embodiments herein disclose a method and system for domain knowledge augmented multi-head attention based robust universal lesion detection. The method utilizes minimal number of Computer Tomography (CT) scan slices to extract maximum information using organ agnostic HU windows and a convolution augmented attention module for a computationally efficient ULD with enhanced prediction performance.

Method and system for domain knowledge augmented multi-head attention based robust universal lesion detection

This disclosure relates generally to intelligent visual reasoning over graphical illustrations using a MAC unit. Prior arts use visual attention to map particular words in a question to specific areas in an image to memorize the corresponding answers, thereby resulting in a limited capability to answer questions of a specific type. The present disclosure incorporates the MAC unit to enable reasoning capabilities and accordingly attend to an area in the image to find the answer. The present disclosure therefore allows generalizing over a possible set of questions with varying complexities so that an unseen question can also be answered correctly based on the reasoning methods that it has learned. The system and method of the present disclosure can be used for understanding of visual information when processing documents like business reports, research papers, consensus reports etc. containing charts and reduce the time spent in manual analysis.

Intelligent visual reasoning over graphical illustrations using a MAC unit

Traditional systems that enable extracting information from Piping and Instrumentation Diagrams (P&IDs) lack accuracy due to existing noise in the images or require a significant volume of annotated symbols for training if deep learning models that provide good accuracy are utilized. Conventional few-shot/one-shot learning approaches require a significant number of training tasks for meta-training prior. The present disclosure provides a method and system that utilizes the one-shot learning approach that enables symbol recognition using a single instance per symbol class which is represented as a graph with points (pixels) sampled along the boundaries of different symbols present in the P&ID and subsequently, utilizes a Graph Convolutional Neural Network (GCNN) or a GCNN appended to a Convolutional Neural Network (CNN) for symbol classification. Accordingly, given a clean symbol image for each symbol class, all instances of the symbol class may be recognized from noisy and crowded P&IDs.

Symbol recognition from raster images of PandIDs using a single instance per symbol class

State of art techniques that utilize spatial association based Table structure Recognition (TSR) have limitation in selecting minimal but most informative word pairs to generate digital table representation. Embodiments herein provide a method and system for TSR from an table image via deep spatial association of words using optimal number of word pairs, analyzed by a single classifier to determine word association. The optimal number of word pairs are identified by utilizing immediate left neighbors and immediate top neighbors approach followed redundant word pair elimination, thus enabling accurate capture of structural feature of even complex table images via minimal word pairs. The reduced number of word pairs in combination with the single classifier trained to determine the word associations into classes comprising as same cell, same row, same column and unrelated, provides TSR pipeline with reduced computational complexity, consuming less resources still generating more accurate digital representation of complex tables.

Method and system for table structure recognition via deep spatial association of words

The need for extracting information trapped in unstructured document images is becoming more acute. A major hurdle to this objective is that these images often contain information in the form of tables and extracting data from tabular sub-images presents a unique set of challenges. Embodiments of the present disclosure provide systems and methods that implement a deep learning network for both table detection and structure recognition, wherein interdependence between table detection and table structure recognition are exploited to segment out the table and column regions. This is followed by semantic rule-based row extraction from the identified tabular sub-regions.

Deep learning based table detection and associated data extraction from scanned image documents

Systems and methods for automating information extraction from piping and instrumentation diagrams is provided. Traditional systems and methods do not provide for end-to-end and automated data extraction from the piping and instrumentation diagrams. The method disclosed provides for automatic generation of end-to-end information from piping and instrumentation diagrams by detecting, via one or more hardware processors, a plurality of components from one or more piping and instrumentation diagrams by implementing one or more image processing and deep learning techniques; associating, via an association module, each of the detected plurality of components by implementing a Euclidean Distance technique; and generating, based upon each of the associated plurality of components, a plurality of tree-shaped data structures by implementing a structuring technique, wherein each of the plurality of tree-shaped data structures capture a process flow of pipeline schematics corresponding to the one or more piping and instrumentation diagrams.

Systems and methods for automating information extraction from piping and instrumentation diagrams

Various methods are using SQL based data extraction for extracting relevant information from images. These are rule based methods of generating SQL-Query from NL, if any new English sentences are to be handled then manual intervention is required. Further becomes difficult for non-technical user. A system and method for extracting relevant from the images using a conversational interface and database querying have been provided. The system eliminates noisy effects, identifying the type of documents and detect various entities for diagrams. Further a schema is designed which allows an easy to understand abstraction of the entities detected by the deep vision models and the relationships between them. Relevant information and fields can then be extracted from the document by writing SQL queries on top of the relationship tables. A natural language based interface is added so that a non-technical user, specifying the queries in natural language, can fetch the information effortlessly.

Method and system for information extraction from document images using conversational interface and database querying

This disclosure relates generally to document processing, and more particularly to extracting information from hand-marked industrial inspection sheets. In an embodiment, the system performs localization of text as well as arrows in the inspection sheet, and identifies text that matches each arrow. Further by identifying machine zone each arrow is pointing to, the system assigns corresponding text to the appropriate machine zone; thus facilitating digitization of the inspection sheets.

Method and system for extracting information from hand-marked industrial inspection sheets

Method and system for automatic chromosome classification is disclosed. The system, alternatively referred as a Residual Convolutional Recurrent Attention Neural Network (Res-CRANN), utilizes property of band sequence of chromosome bands for chromosome classification. The Res-CRANN is end-to-end trainable system, in which a sequence of feature vectors are extracted from the feature maps produced by convolutional layers of a Residual neural networks (ResNet), wherein the feature vectors correspond to visual features representing chromosome bands in an chromosome image. The sequence feature vectors are fed into Recurrent Neural Networks (RNN) augmented with an attention mechanism. The RNN learns the sequence of feature vectors and the attention module concentrates on a plurality of Regions-of-interest (ROIs) of the sequence of feature vectors, wherein the ROIs are specific to a class label of chromosomes. The Res-CRANN provides higher classification accuracy as compared to the state-of the-art methods for chromosome classification.

Method and system for automatic chromosome classification

The most challenging problems in karyotyping are segmentation and classification of overlapping chromosomes in metaphase spread images. Often chromosomes are bent in different directions with varying degrees of bend. Tediousness and time consuming nature of the effort for ground truth creation makes it difficult to scale the ground truth for training phase. The present disclosure provides an end-to-end solution that reduces the cognitive burden of segmenting and karyotyping chromosomes. Dependency on experts is reduced by employing crowdsourcing while simultaneously addressing the issues associated with crowdsourcing. Identified segments through crowdsourcing are pre-processed to improve classification achieved by employing deep convolutional network (CNN).

Crowdsourcing and deep learning based segmenting and karyotyping of chromosomes

This disclosure relates generally to character detection and recognition, and more particularly to a method and system for container code recognition via Spatial Transformer Networks and Connected Component. The method comprises capturing an image of a container using an image capture device which is pre-processed using an image preprocessing module. The method further comprises extracting and filtering region proposals from the pre-processed image using a region extraction module to generate regrouped region proposals. The next step comprises classifying the regrouped region proposals into characters by implementing trained Spatial Transformation Network to generate a valid group of region proposal with more than one chunk of container identification code using a classification module, and lastly a sequence for the valid group of region proposal is generated and the generated sequence is mapped to a predefined standard container identification code to determine a container identification code, wherein the predefined standard identification code comprises chunks of characters in a predefined pattern.

Method and system for container code recognition

This disclosure relates generally to image processing, and more particularly to image texture determination. In one embodiment a processor a memory coupled to the processor, wherein the processor coupled with a plurality of modules stored in the memory: At least one image having a plurality of pixels is accepted. Any noise is removed from the image to obtain at least one noise free image. The at least one noise free image is converted to at least one gray scale image. Horizontal and Vertical Gradient for plurality of pixels of the at least one gray scale image are computed. Gradient magnitude is calculated for the generated gradient. Histogram of the gradient magnitude is generated based on the gradient magnitude, and the plurality of generated histograms are compared with a plurality of predetermined histograms.

Computer implemented system and method for image texture determination

A method and system is provided for geo-demographic classification of a geographical region. The present application discloses an unsupervised learning method and system for analyzing satellite imagery and multimodal sensory data in fusion for geo-demographic clustering. The present application also discloses an inexpensive and faster method and system for geo-demographic classification of a geographical region.

Growing community of inventors

Gurgaon, India

Monika Sharma