An AI system for the Social Thermometer, a way to evaluate the risk temperature of the places by using webcam live streams.
We need safety and social distance if we are to restart the lives of our society. Law enforcement wants to protect us but they cannot be everywhere where they are needed to attend the needs of the general public. Additionally, people want their mobility to be safe and will not want to be at a place where the presence of other people may compromise public health. We want to help both the general public and law enforcement with the knowledge of if a certain location is safe or if it is a potential health hazard.
Crowded places can be risky to go in the new reality that will come after the COVID-19 and who knows what other kinds of spreading disease that might appear. Social Thermometer could help everybody to make better plans and check upon social distancing policies.
We were all living our daily lives, but then the COVID 19 happened. Now we all shall stay home to save the ones we love. Some of our favorite places are now closed or under restricted control rules and we won't get back to it in a while. The problem is: How can we speed up this process? Social distancing seems a good way, but how to ensure the risk of going out is low enough?
The solution is SafeSpot! A system that uses technology to create a social distancing thermometer. Recent studies have already established a direct relationship between the infectious potential with crowdedness.
The risk degrees are calculated by the number of proximity incidents over time. Crowded places have bigger probabilities of social contact, touching, and infection, which means they are risky. And we all prefer cooler places!
The SafeSpot system uses state-of-the-art artificial intelligence algorithms to perform video analysis over any kind of webcams, either using the public ones already available either your stream! The result is an automatic social distancing analysis tool that allows us to compare places by their social risk temperature.
We use feeds of public live cams that are placed everywhere (beaches, public amenities, parks, roads) and analyze it, in real-time to determine if their locations are secure, from a social distancing point of view, meaning that the stakeholders can know, in real-time, if at a certain location there is a social gathering that may be compromising public health. The security personal will have real-time information about risky situations and preventive policies enforcement.
Development Log during the Hackathon
In these three days of the hackathon, we pushed all our efforts to develop a functional prototype you can try on our website. The use cases of SafeSpot are unlimited, but we want to showcase two of our main scenarios.
Day One (April, 24):
- Team gathering
- Problem definition, goals, strategy and solution proposal
- Roles definition
- Proposal submission
- Development start
Day Two (April, 25):
- Software development:
- Python script for getting video streams from URL
- Training Artificial Intelligence Model for people detection
- Definition of risk metrics
- Webpage design and deployment
- Business model definition
Day Three (April, 26):
- Software development
- Risk evaluation scripts
- Disseminate project through social networks
- Pitch Preparation
- Submission finalization
Showcase Application Scenarios
We have designed our solution envisaging multiple application scenarios, some that are more urgent and should provide general public and authorities help in this COVID-19 crisis and others that are more business-oriented.
Scenario one regards private business partnership. Consider you want to go out and have dinner with your family, but you don't know if the restaurants are safe enough. SafeSpot could help you find out the cool places around the town and decide. Besides that, our system would help owners to provide governments enough evidence they are implementing preventive measures.
In the second scenario, you want to go out to walk around, but don't know how to get safe to that beautiful park. While you could check out and create a schedule to avoid crowds, authorities, traffic managers, and the police could use SafeSpot to reinforce control on the hotspots!
Impact to the COVID-19 Crisis and beyond
Hot and Cool classification mechanism would help everybody during the COVID 19 crisis and generate risk data over time so we all can be prepared to face future events like this. We firmly believe in the future of SafeSpot, because it is ready to deploy, low cost and it could be used for many other applications in all sectors, always respecting the privacy of the EU citizens and trusting on open-source code.
SafeSpot is the way to get life back on track!
Value of the solution
SafeSpot uses existing resources (live cam streams) and respects the privacy of EU citizens to provide valuable insights about the risk of infection in public and private places:
- By using the birds-eye cameras already public (it’s free!)
- By authorized personnel using smart-glasses and first-person view
- Evaluate and compare risk in different environments such as restaurants, supermarkets, etc. to avoid queues and turmoil
- Useful to analyze the data and compare crowd’s movements in all these situations, so we could predict all risky situations in the future
We have also developed a business model for SafeSpot during the weekend:
Challenges and Future work
Short-term challenges include:
- Get in touch with authorities and stakeholders
- Dissemination efforts
- Computational power to run at scale
Long-term challenges include:
- Partnership with governments and business