Growing population, growing vehicles and lack of awareness has led to the problem of accidents happening at low to high density vehicle and human population. A witness in an accident has no medical training and often becomes a bystander. Drones as emergency responders exploits the use of ICT with AI to as an add-on to emergency services. With a push of a button by a witness, the drones use the GPS of the phone to reach the spot, the AI model is trained to classify the accident and shares its results with the emergency responders including hospitals. The ambulance reaching the spot is prepared with the basic of what has to be done and the nearest hospital well prepared to act.

 
 

Outline/Structure of the Talk

To bring in focus of using hardware and software into healthcare. Drones and AI in healthcare especially as emergency responders will save a lot of lives and change the course of how emergency situations are handled. Drones trained to identify what kind of injuries has occurred and relaying information to emergency responders will help in being better prepared and organised.

Accident Images Data set:

  1. Accident detection: Images labelled with accident and images labelled without accident.
  2. Vehicles in accident:Images labelled with Light/heavy/motorcycle
  3. Accident Level- Images labelled :high/medium/low

Type of injury data set:

  1. Head Injury: Images labelled head injury and non-head injury
  2. Torso Injury: Images labelled torso injury and non-torso injury
  3. Lower Body Injury- Images labelled lower body injury and non-lower body injury.

All trained using CNN

For drones the best way was to use K-Nearest Neighbour (KNN)

Difficulties:

Apart from using just images, the use of other pointers such as body temperature , hear beat detection, blood clot level/ % of blood clot, dislocated joints/broken bones would make the model more versatile.

One of the major difficulty is to remove the noise from the images which may be due to pixelation or debris in the images.

Breakdown of the talk:

  • Introduction: 2 min
  • The requirement/necessity of the solution: 2 min
  • The how of the solution: 2 min
  • Types of data sets, training and the steps : 7 min
  • Amalgamation of accident trained module with drone trained in KNN: 2 minutes
  • Application of solution:1 min
  • Conclusion: 2 min
  • Questions: 2 min

Learning Outcome

  • Looking at helping the society.
  • Generating awareness on saving the golden minute during accidents.
  • Creating awareness that Drones can be used in Healthcare.
  • Basic awareness about AI application
  • How to build modules

Target Audience

Early Thinkers, AI enthusiasts, Drone developers, ML/Deep Learning enthusiasts

Prerequisites for Attendees

Basics of ML/DL/AI, and an innovative mind

Slides


Video


schedule Submitted 1 year ago

Public Feedback


    • Amarjeet Dua
      keyboard_arrow_down

      Amarjeet Dua - Case Study: Fusing Machine Learning into Operations Research Techniques to solve Complex Optimization Problems

      20 Mins
      Experience Report
      Intermediate

      Machine Learning (ML) and Operations Research (OR) have co-existed for long. There have been amazing applications driven by ML and OR that we come across in our day-to-day lives. These applications range from matching algorithms on dating websites to solving large scale vehicle routing problems for complex supply chains. But, have you ever wondered what happens when these two areas of mathematical science come together to solve complex real-world optimization problems?

      Are you curious to know how OR-applications can benefit from the power of ML?

      In this talk, we’ll go through a real-world case study where we used the power of (ML+OR) to create significant dollar savings in the area of airline flight schedules. I will also take you through cases where ML can help OR solutions shine further to solve more generic problems.

    help