OptiSol Business Solutions
Nominated Award: Best Application of AI to achieve Social Good
Website of Company (or Linkedin profile of Person): https://www.optisolbusiness.com/
We are a global tech company with 200+ clients and 5 global offices including Ireland with 400+ happy employees. We have been in operations for about 13 years now and were awarded by reputed firms like Clutch, CII, NASSCOM, CIO, Express Media and we are select partners with AWS and Microsoft. Organizations are facing challenges with digital transformation and adapting technology changes and we help them as a trusted tech modernization partner with a proven engagement model to make their business processes more agile and innovative and make them a future ready digital organization. As a global tech company with clients across globe and business verticals, we are helping them in their digital transformation journey by implementing latest technologies such as AI and ML, IoT and DataOps. We collaborate with clients to digitally transform their business, improve productivity, enhance customer experience and reduce cost.
Reason for Nomination:
What is the problem you are trying to solve?
Workplace injury is an event or exposure in the work environment caused or contributed to the injury or significantly aggravated a pre-existing injury. Musculoskeletal disorders are the costliest and frequent workplace injuries. Musculoskeletal disorders are injuries or disorders of the muscles, nerves, tendons, joints, cartilage, and spinal discs, caused due to improper lifting of heaving objects.
Musculoskeletal disorders account for nearly 70 million physician office visits in the United States annually, and an estimated 130 million total health care encounters including outpatient, hospital, and emergency room visits. In 1999, nearly 1 million people took time away from work to treat and recover from work-related musculoskeletal pain or impairment of function in the low back or upper extremities.
We believed that latest advances in Computer Vision and Machine Learning can be leveraged here to build a mobile app that can automate Ergonomics assessment. This will in turn reduce repetitive stress injuries at work place and prevent serious health issues. We went ahead and built our Posemedic app. A demo of this app is show in this YouTube video
https://www.youtube.com/watch?v=dlZrG4e-swE&feature=youtu.be
How did you go about it?
Our Preventive Healthcare solution uses the Multi-angular Pose estimation technique for tracking the posture movement of worker’s key points while they are lifting heavy objects.The key points denotes the spatial location of major joints like elbow, knee etc. This computer vision solution uses Pose Estimation and Rapid Upper Limb Assessment (RULA) score to evaluate the worker’s posture movement and determines whether the upper extremity task leads to Musculoskeletal problems. Based on the postural angle of the different body posture, the RULA score will be determined and shows the risk factors based on the action level output.The solution considers not only the upper pose angle but also the side pose angles to get accurate determination of the posture movement.
Our Posemedic app provides a preventive healthcare methodology to screen workers who have a significant risk of getting musculoskeletal disorders by tracking their posture movements through android devices. In case a worker tries to carry a heavy object with improper posture movement ,the actual output score shows a risk of MSD and sends a warning signal immediately. Based on the angles of different body positions (upper arm, lower arm, wrist, wrist twist, muscle use score, force/load score, Low neck, trunk position, legs) RULA score is calculated. The consideration of multiple angles of body posture (front, & side poses) for determining end score improves the accuracy of providing the greatest exposure of MSD. This application advantage includes minimal time effort and equipment for determining risk factors in workplace.
What technologies did you use and why?
Our app is developed using Kotlin language, which is a cross-platform, statically typed, general-purpose programming language with type inference. We integrated a TensorFlow lite model named Movenet Thunder to perform pose estimation on the video frames from the mobile phone camera. This TensorFlow lite model is an optimized model suitable for running in mobile phones with their limited processing power. This model generates the co-ordinates of the key points in the human body, which we convert into body angles and generate the RULA scores from it.
How difficult was it and the challenges?
The working of this project is based on calculating angles between body postures. We couldn’t determine every angle at a given pose for an individual. So, we will check if the person is facing straight or he is facing sidewards and calculate the angles accordingly. The next challenge is that our project pipeline is designed for only one person to be in the frame at a time. Even when more than one person is there, we just take one person with maximum confidence for calculation.
What was the result?
Provides personalized interventions about improper posture movements that may cause musculoskeletal disorders. Helps companies in avoiding spending of exorbitant costs in paying fines and compensation claims. Company considerations in safeguarding workers’ health improves Employer-Employee relationship. Avoids frequent sprains and overexertion injuries during work hours that it reduces employee’s time offs.