Abstract: Disclosed herein is a system for identifying root causes of industrial ergonomic risks and their corresponding solutions. An example system comprises a computing device that trains a deep learning based model to process an embedding vector for outputting a root cause of an ergonomic risk identified in an image and a solution to the ergonomic risk. A dataset for training the model is established by collecting samples of image-caption data that includes a plurality of images and information relating to ergonomic problems identified in the plurality of images and corresponding solutions to the ergonomic problems.

Abstract: Disclosed herein are a vision-based three-dimensional (3D) pose estimation system and a method for ergonomic risk assessment. An example system comprises a computing device configured to obtain a video capturing motions of a subject performing at least one working activity for a selected duration of time, perform a whole-body two dimensional (2D) pose estimation based at least on extracted frames of the video, perform a whole-body 3D pose estimation based at least on the whole-body 2D pose estimation, calculate joint angles based at least on the whole-body 3D pose estimation, and determine an ergonomic risk level of each identified joint using the joint angles.

Abstract: A computing system and method for identifying potential serious injuries and fatalities (PSIF) incidents from incident records. An example system may include a computing device configured to obtain an incident record, identify fields included in the incident record in connection with a set of descriptors, generate a textual description of the incident record based on selected fields of the incident record corresponding to the set of descriptors, determine parameters and natural language instructions for using a language model to determine whether the incident record is a PSIF incident, instruct the language model to perform a machine-learning based binary classification to generate a PSIF or non-PSIF label and a rationale for the incident record based upon the parameters and natural language instructions and the textual description of the incident record, and generate data for a display device to indicate the PSIF or non-PSIF label and the rationale thereof.

Abstract: A system is disclosed for identifying industrial ergonomics risk root-causes and providing risk control actions. The system includes a processor configured to execute machine readable instructions to receive, from a computing device of a user, video of an individual performing a task. The processor is also configured to process the video of the individual performing the task through a deep learning framework to determine ergonomic risk root-causes. The processor is further configured to determine ergonomic risk control recommendations to mitigate the determined ergonomic risk root-causes. The processor transmits, to the computing device, information indicative of the ergonomic risk control recommendations and the ergonomic risk root-causes.

Abstract: Disclosed herein is a machine learning based system and method for root cause identification and corrective action recommendation to address various workplace incidents. An example system may include a computing device configured to formulate a root cause superset to represent a plurality of root causes, identify fields included in a received incident record in connection with a set of predetermined field descriptors to generate a textual description of the incident record, store data relating to the root cause superset and the textual description of the incident record as a trained model, instruct LLM(s) to determine at least one root cause and corrective action based least upon the textual description of the incident record and the root cause superset, and generate data for a display device to indicate the at least one root cause and corrective action.

Abstract: Disclosed herein is an image-based system configured to identify root causes of industrial ergonomic risks and their corresponding solutions. An example system comprises a computing device configured to encode an image of a worker performing a work task to generate an embedding vector, transmit the embedding vector to an image-grounded text decoder, while generating first tokens to instruct the decoder to generate a first sentence indicating a root cause of an ergonomic risk identified in the image, compute first relative sensitivity scores relating to the first tokens and extracted image features, generate second tokens of the first sentence based on the first relative sensitivity scores, while generating third tokens to instruct a text decoder to generate a second sentence indicating a solution to the ergonomic risk, calculate second relative sensitivity scores relating to the second and third tokens, and generate the first and second sentences accordingly.

Abstract: A computing system and method are configured to receive an input in a portable document format, generate a raster image of the input, and detect a form type of the input based on the raster image. In response to detecting that the raster image of the input indicates a first form type, the computing system and method are configured to partition the raster image into sections corresponding to document regions containing document entities targeted for retrieval for a specific form type, generate sub-images based on bounding coordinates of the sections, and apply selected document data identification and retrieval computational techniques to extract the document entities from the sub-images.

Abstract: A system is disclosed for identifying industrial ergonomics risk root-causes and providing risk control actions. The system includes a processor configured to execute machine readable instructions to receive, from a computing device of a user, video of an individual performing a task and receive, from the computing device, textual information related to the task. The processor is also configured to process the textual information and the video of the individual performing the task through a deep learning framework to determine ergonomic risk root-causes. The processor is further configured to determine ergonomic risk control recommendations to mitigate the determined ergonomic risk root-causes. The processor transmits, to the computing device, information indicative of the ergonomic risk control recommendations and the ergonomic risk root-causes.

Abstract: Disclosed herein are vision-based three-dimensional (3D) pose estimation system and method for ergonomic risk assessment. An example system may comprise a computing device configured to obtain a monocular video capturing motions of a subject performing at least one working activity for a selected duration of time, perform a whole-body two dimensional (2D) pose estimation based at least on extracted frames of the monocular video, perform a whole-body 3D pose estimation based at least on the whole-body 2D pose estimation, calculate joint angles based at least on the whole-body 3D pose estimation, determine a posture score for each identified joint in each frame of the monocular video, and determine an ergonomic risk level of each identified joint based at least upon the posture score.

Abstract: Computing systems and methods are configured to structure unstructured document data. An example computing system is configured to receive an input in the form of a document containing unstructured data, determine a plurality of entities to be identified and retrieved from the input, analyze each entity to determine identification and retrieval computational techniques, group the plurality of entities into different categories based on the identification and retrieval computational techniques, generate multiple images of the input in different dot per inches, extract machine-readable texts from the multiple images of the input, conduct a text extraction for native PDF in the input, apply the identification and retrieval computational techniques to identify and retrieve the plurality of entities in different categories based at least upon the machine-readable texts extracted from the multiple images of the input and the native PDF in the input, and save processing results in at least one structured database.

Abstract: A system for identifying industrial ergonomics risk root-causes and providing risk control actions, comprising: a computing device configured to obtain video signals of a worker performing a job at a workplace; and a computing server system configured to receive and process the video signals to determine joint locations of the worker, calculate joint angles for each body region of the worker based on the joint locations, calculate, based on the joint angles, a risk score for each body region of the worker in a plurality of risk categories, calculate a risk rating for each body region of the worker based on the risk score, determine ergonomic risk root-causes for each body region of the worker based on the risk rating, and provide ergonomic risk control recommendations to mitigate the ergonomic risk root-causes.

Abstract: Disclosed herein is a system and method for obtaining and generating motion capture datasets relating to various working activities for ergonomic risk assessment. An example system may comprise a computing device that obtains first data from multiple motion capture cameras, obtains second data from multiple visible light imaging sensors, calculates 3D positions of multiple reflective markers positioned on several subjects performing various working activities based on the first data, labels each marker to generate marker trajectories, performs gap filing and smoothing functions on the marker trajectories to generate global marker positions, transforms the global marker positions into a corresponding image coordinate system of each sensor to generate 3D pose data of the subjects at each sensor viewpoint, projects the 3D pose data into frames of the second data to generate 2D pose data, and generates a dataset comprising the second data, the 2D pose data, and the 3D pose data.

Abstract: A system, comprising: a computing device configured to obtain video signals of a worker performing a hand-related job at a workplace; and a computing server system configured to receive and process the video signals to identify hand grips and wrist bending involved in the job, determine a hand grip type for each identified hand grip, obtain force information relating to each identified hand grip, determine neutral or hazardous wrist bending based at least upon the wrist bending and the hand grip force information, calculate a percent maximum strength for each identified hand grip, calculate frequencies and durations of each identified hand grip and wrist bending, and determine ergonomic risks of the hand-related job accordingly.

Abstract: Systems and methods to classify incident report documents are disclosed, comprising inputting, a first type data entry of a document into a deep neural network (DNN); encoding, via the DNN, the first type data entry to output a densely embedded contextual vector representing contents of the first type data entry; generating, a list containing ordered data from a second type data entry of the document; encoding, via a machine learning network, the ordered data into a sparse vector representation of the second type data entry; concatenating, the densely embedded contextual vector with the sparse vector representation to generate a representative vector of the document; and training a gradient-boosted classifier network by using as training inputs the representative vector and a label associated with the document to generate a classification of the document.

Abstract: Disclosed herein is an image-based system configured to identify root causes of industrial ergonomic risks and their corresponding solutions. An example system comprises a computing device configured to encode an image of a worker performing a work task to generate an embedding vector, transmit the embedding vector to an image-grounded text decoder, while generating first tokens to instruct the decoder to generate a first sentence indicating a root cause of an ergonomic risk identified in the image, compute first relative sensitivity scores relating to the first tokens and extracted image features, generate second tokens of the first sentence based on the first relative sensitivity scores, while generating third tokens to instruct a text decoder to generate a second sentence indicating a solution to the ergonomic risk, calculate second relative sensitivity scores relating to the second and third tokens, and generate the first and second sentences accordingly.

Abstract: Systems and methods are disclosed for automated industrial hygiene assessment and display comprising receiving sampling results for a stressor, such as a harmful environmental artifact in a physical environment; deriving one or more codes for the stressor from a digital record via an indexing module and/or from other data sources; generating a health effect rating (HER) based on the code; generating an exposure rating (ER) based on the sampling results; generating an uncertainty rating (UR) based on the sampling results; displaying, an interactive UIto facilitate approval or selection of at least one of the HER, the ER, or the UR; generating at least one of a risk rating (RR) or an information gathering priority rating (IGPR) based on a selection of the at least one of the HER, the ER, or the UR; and displaying via the interactive UI at least one of the RR or the IGPR.

Abstract: A computing device of a system is deployed within a communication network for processing environmental, social, and governance (ESG) reports. The computing device is configured to obtain an ESG report of an entity, receive a question in natural language regarding information contained in or inferable from the ESG report, convert the ESG report into page-wise pieces of text, determine an answer to the question based at least upon the page-wise pieces of text of the ESG report, determine a location of a snippet related to the answer in the page-wise pieces of text of the ESG report, generate a bounding box through which the snippet is highlighted in a page containing the snippet, crop the page based on the bounding box, and display a cropped image of the page.

Abstract: Systems and methods to classify incident report documents are disclosed, comprising inputting, a first type data entry of a document into a deep neural network (DNN); encoding, via the DNN, the first type data entry to output a densely embedded contextual vector representing contents of the first type data entry; generating, a list containing ordered data from a second type data entry of the document; encoding, via a machine learning network, the ordered data into a sparse vector representation of the second type data entry; concatenating, the densely embedded contextual vector with the sparse vector representation to generate a representative vector of the document; and training a gradient-boosted classifier network by using as training inputs the representative vector and a label associated with the document to generate a classification of the document.

Abstract: A system, comprising: a computing device configured to obtain video signals of a worker performing a hand-related job at a workplace; and a computing server system configured to receive and process the video signals to identify hand grips and wrist bending involved in the job, determine a hand grip type for each identified hand grip, obtain force information relating to each identified hand grip, determine neutral or hazardous wrist bending based at least upon the wrist bending and the hand grip force information, calculate a percent maximum strength for each identified hand grip, calculate frequencies and durations of each identified hand grip and wrist bending, and determine ergonomic risks of the hand-related job accordingly.

Abstract: Systems and Methods are disclosed herein for automatically indexing multiple informational fields in digital data records, the method comprising: identifying, based on rules defining target information fields, for each target field of the target information fields, at least one page in a digital data record comprising content related to the target field; extracting, for each target field, from the identified at least one page, at least one portion of text comprising the content; feeding, for each target field, a pre-processed version of the at least one portion of text into a machine learning (ML) model, wherein the ML model is trained on the target field; determining, for each target field, via the ML model trained on the target field, at least one candidate text comprising the content; and extracting, for each target field, the at least one candidate text.