Abstract: This disclosure relates generally to the field of Electroencephalogram (EEG) classification, and, more particularly, to method and system for EEG motor imagery classification. Existing deep learning works employ the sensor-space for EEG graph representations wherein the channels of the EEG are considered as nodes and connection between the nodes are either predefined or are based on certain heuristics. However, these representations are ineffective and fail to accurately capture the underlying brain's functional networks. Embodiments of present disclosure provide a method of training a weighted adjacency matrix and a Graph Neural Network (GNN) to accurately represent the EEG signals. The method also trains a graph, a node, and an edge classifier to perform graph classification (i.e. motor imagery classification), node and edge classification. Thus, representations generated by the GNN can be additionally used for node and edge classification unlike state of the art methods.

Abstract: Model-based image reconstruction (MBIR) methods using convolutional neural networks (CNNs) as priors have demonstrated superior image quality and robustness compared to conventional methods. Studies have explored MBIR combined with supervised and unsupervised denoising techniques for image reconstruction in magnetic resonance imaging (MRI) and positron emission tomography (PET). Unsupervised methods like the deep image prior (DIP) have shown promising results and are less prone to hallucinations. However, since the noisy image is used as a reference, strategies to prevent overfitting are unclear. Recently, Bayesian DIP (BDIP) networks that model uncertainty tend to prevent overfitting without requiring early stopping. However, BDIP has not been studied with data-fidelity term for image reconstruction. Present disclosure provides systems and method that implement a MBIR framework with a modified BDIP.

Abstract: Portable ECG monitors available in market have the disadvantage that the ECG data they provide as input aren't directly interpretable and requires medical knowledge for the users. The disclosure herein generally relates to Electrocardiogram (ECG), and, more particularly, to a method and system for generating 2d representation of electrocardiogram (ECG) signals. The system provides a mechanism for determining variability between a plurality of segments of an ECG data measured, and uses the information on the determined variability to generate the 2D representation corresponding to the ECG signal. The system further provides means to generate a data model that can be further used for processing real-time ECG data for generating corresponding interpretations. This allows a user to obtain the interpretations as output.

Abstract: The disclosure generally relates to scene graph generation. Scene graph captures rich semantic information of an image by representing objects and their relationships as nodes and edges of a graph and has several applications including image retrieval, action recognition, visual question answering, autonomous driving, robotics. However, to leverage scene graphs, computationally efficient scene graph generation methods are required, which is very challenging to generate due presence of a quadratic number of potential edges and computationally intensive/non-scalable techniques for detecting the relationship between each object pair using the traditional approach. The disclosure proposes a combination of edge proposal neural network and the Graph neural network with spatial message passing (GNN-SMP) along with several techniques including a feature extraction technique, object detection technique, un-labelled graph generation technique and a scene graph generation technique to generate scene graphs.

Abstract: This disclosure relates generally to identification and mitigation of bias while training deep learning models. Conventional methods do not provide effective methods for bias identification, and they require pre-defined concepts and rules for bias mitigation. The embodiments of the present disclosure train an auto-encoder to produce a generalized representation of an input image by decomposing into a set of latent embedding. The set of latent embedding are used to learn the shape and color concepts of the input image. The feature specialization is done by training an auto-encoder to reconstruct the input image using the shape embedding modulated by color embedding. To identify the bias, permutation invariant neural network is trained for classification task and attribution scores corresponding to each concept embedding are computed. The method also performs de-biasing the classifier by training it with a set of counterfactual images generated by modifying the latent embedding learned by the auto-encoder.

Abstract: This disclosure relates generally to methods and systems for automated image segmentation of an anatomical structure such as heart. Most of the techniques in literature are using 2-D or slice by-slice data due to lightweight and need of less data for training. These networks lack 3-D contextual information. Further, the conventional techniques are inaccurate and inefficient in the 3-D image segmentation till the last slice of the image. The present disclosure solves automated 3-D image segmentation of the anatomical structure such as heart, by proposing a new Generative Adversarial Network (GAN) based architecture for the 3-D segmentation, with a patch-based extraction technique and a class-weighted generalized dice loss. The proposed 3-D GAN based architecture is capable of storing the 3-D contextual information for the image segmentation of the anatomical structure, with high accuracy.

Abstract: State of art techniques for color regeneration are complex and fail to provide color control. Embodiments of the present disclosure provide a method and system for generating color variants for fashion apparels by providing a Fashion Apparel Regeneration-Generative Adversarial Network (FAR-GAN) to generate color variants of the fashion apparels. The FAR-GAN utilizes a two-step encoding process to encapsulate both an input image and an edge-map information along with a target color embedding branch to manipulate the color information present in the fashion apparel present in the input image that is to be changed to a desired target color. Furthermore, the color and structural information is disentangled by controlling them using a color consistency loss. The FAR-GAN can be trained end-to-end without incorporating complex multi-step process.

Abstract: Optical images in remote sensing are contaminated by cloud cover and bad weather conditions and are only available during the daytime. Whereas SAR images are completely cloud free, independent of weather conditions and can be acquired both during the day and at night. However, due to the speckle effect and side looking imaging mechanism of SAR images, they are not easily interpretable by untrained people. To address this issue, the present disclosure provides a method and system for LULC guided SAR visualization, wherein a GAN is trained to translate SAR images to optical images for visualization. A given SAR image is fed into a first generator of the GAN to obtain LULC map which is then concatenated with the SAR image and fed into a second generator of the GAN to generate an optical image. The LULC map provides semantic information required for generation of more realistic optical image.

Abstract: Conventional task planners assume that the task-plans provided are executable, hence these are not task-aware. Present disclosure alleviates the downward refinability assumption, that is, planning can be decomposed separate symbolic and continuous planning steps by introducing bi-level planning, a plan which is a series of actions that the robot needs to take to achieve the goal task is curated. Firstly, abstract symbolic actions are converted to continuous vectors and used therein to enable interaction with an environment. Images of objects placed in the environment are captured and concepts are learnt from the captured images and attributes of objects are detected. A hierarchical scene graph is generated from the concepts and attributes wherein the graph includes interpretable sub-symbolic representations and from these interpretable symbolic representations are obtained for identifying goal task.

Abstract: This disclosure relates generally to systems and methods for autonomous task composition of vision pipelines using an algorithm selection framework. The framework leverages transformer architecture along with deep reinforcement learning techniques to search an algorithmic space for unseen solution templates. In an embodiment, the present disclosure describes a two stage process of identifying the vision pipeline for a particular task. At first stage, a high-level sequence of the vision pipeline is provided by a symbolic planner to create the vision workflow. At second stage, suitable algorithms for each high-level task are selected. This is achieved by performing a graph search using a transformer architecture over an algorithmic space on each component of generated workflow. In order to make the system more robust, weights of embedding, key and query networks of a visual transformer are updated with a Deep Reinforcement Learning framework that uses Proximal Policy Optimization (PPO) as underlying algorithm.

Abstract: Current inspection processes employed for pipeline networks data acquisition aided with manually locating and recording defects/observations, thus leading labor intensive, prone to error and a time-consuming task thereby resulting in process inefficiencies. Embodiments of the present disclosure provide systems and methods for that leverage artificial intelligence/machine learning models and image processing techniques to automate log and data processing, reports and insights generation thereby reduce dependency on manual analysis, improve annual productivity of survey meterage and bring in process and cost efficiencies into overall asset health management for utilities, thereby enhancing accuracy in defect identification, analysis, classification thereof.

Abstract: State of the art mechanisms being used for achieving diagnostic-quality images under low-dose settings for general CT imaging have the disadvantages that CT images are fixed during the optimization process to generate perfusion maps, which can lead to suboptimal CT images with respect to the perfusion maps generated, although they might appear spatially smooth or denoised. The disclosure herein generally relates to Computer Tomography (CT) scanning, and, more particularly, to a method and system for CT image reconstruction. The system performs modelling an optimization problem for joint estimation of a set of structural CT images and a perfusion map, and further solves the optimization problem for the reconstruction of the CT images of a subject.

Abstract: State of the art systems used for airport automation and data processing may be prone to data security related issues, as unauthorized personal may gain entry to sensitive data. The disclosure herein generally relates to airport management, and, more particularly, to a method and system for service authentication in an airport management network. The system uses a neural network to process a received service request and decides whether the service request is to be allowed or denied, based on a determined validity of the service request, role based access defined for a user requesting the service, a feature map data generated.

Abstract: State of the art systems being used for QSM reconstruction have explored prior information from both magnitude and phase data. However, the underlying assumption is that the susceptibility maps and the magnitude images have coinciding edges. Establishing the ground-truth susceptibility maps is difficult and leads to limited applicability of supervised methods. Further, with portable MRI machines becoming a reality, low-field imaging is getting more prominence, which brings in several associated challenges due to noise and external interference. The disclosure herein generally relates to magnetic resonance imaging (MRI) imaging systems, and, more particularly, to a method and system for quantitative susceptibility mapping (QSM) reconstruction in magnetic resonance imaging (MRI) systems. The system performs an iterative reconstruction of QSM, wherein in each iteration the reconstructed QSM from previous iteration is refined by comparing with a reference image generated using same subject's prior MRI data.

Abstract: Conventionally, systems and methods have been provided for manual annotation of anatomical landmarks in digital radiography (DR) images. Embodiments of the present disclosure provides system and method for anatomical landmark detection and identification from DR images containing severe skeletal deformations. More specifically, motion artefacts and exposure are filtered from an input DR image to obtain a pre-processed DR image and probable/candidate anatomical landmarks comprised therein are identified. These probable candidate anatomical landmarks are assigned a score. A subset of the candidate anatomical landmarks (CALs) is selected as accurate anatomical landmarks based on comparison of the score with a pre-defined threshold performed by a trained classifier. Position of remaining CALs may be fine-tuned for classification thereof as accurate anatomical landmarks or missing anatomical landmarks.

Abstract: This relates generally to a method and a system for spatio-temporal polarization video analysis. The spatio-temporal polarization data is analyzed for a computer vision application such as object detection, image classification, image captioning, image reconstruction or image inpainting, face recognition and action recognition. Numerous classical and deep learning methods have been applied on polarimetric data for polarimetric imaging analysis, however, the available pre-trained models may not be directly suitable on polarization data, as polarimetric data is more complex. Further compared to analysis of the polarimetric images, a significant number of actions can be detected by polarimetric videos, hence analyzing polarimetric videos is more efficient. The disclosure is a spatio-temporal analysis of polarization video.

Abstract: Existing Mueller Matrix polarization techniques that rely only on polarization properties are insufficient for accurate characterization of transparent objects. Embodiments of the present disclosure provide a method and system for Mueller Matrix polarimetric characterization of transparent object using optical properties along with the polarization properties to accurately characterize the transparent object. The polarization properties of are derived from a decomposed Mueller matrix element. Additionally, the method derives the optical properties by further subjecting the decomposed Mueller matrix element to Fresnel’s law-based analysis and a reverse Monte Carlo analysis to extract optical properties such as a material refractive index and a material attenuation index. Optical properties vary with changes in the material property caused due to several factors such as manufacturing defect, aberration, inclusion of an impurity such as bubble or dust etc.

Abstract: State of the art techniques in the domain of video analysis have limitations in terms of capability to extract the spatial and temporal information. This limitation in turn affects interpretation of the video data. The disclosure herein generally relates to video analysis, and, more particularly, to a method and system for video analysis to extract spatio-temporal information from a video being analyzed. The system uses a neural network architecture which has multiple layers to extract spatial and temporal information from the video being analyzed. The method of training the neural network that extracts a micro-scale information from a latent representation of the video is presented. This is generated using an attention network, which is then used to extract spatio-temporal information corresponding to the collected video, which is then used in multiple video analysis applications such as searching actions in videos, action detection and localization.

Abstract: State of the art techniques in the domain of video analysis have limitations in terms of capability to capture spatio-temporal representation. This limitation in turn affects interpretation of video data. The disclosure herein generally relates to video analysis, and, more particularly, to a method and system for video analysis to capture spatio-temporal representation for video reconstruction and analysis. The method presents different architecture variations using three main deep network components: 2D convolution units, 3D convolution units and long short-term memory (LSTM) units for video reconstruction and analysis. These variations are trained for learning the spatio-temporal representation of the videos in order to generate a pre-trained video analysis module. By understanding the advantages and disadvantages of different architectural configurations, a novel architecture is designed for video reconstruction.

Abstract: Traditional systems used for fashion attribute detection struggle to generate accurate predictions due to presence of large intra-class and relatively small inter-class variations in data related to the fashion attributes. The disclosure herein generally relates to image processing, and, more particularly, to a method and system for fashion attribute detection. The method proposes F-AttNet, an attribute extraction network to leverage the performance of fine-grained localized fashion attribute recognition. F-AttNet comprises Attentive Multi-scale Feature Encoder (AMF) blocks that encapsulate multi-scale fine-grained attribute information upon adaptive recalibration of channel weights. F-AttNet is designed by hierarchically stacking the AMF encoders to extract deep fine-grained information across multiple scales.