Co-Video as a response to support safe lockdown removal and business activities recovery

Following COVID19 and its lockdown measures individuals, societies, and economies are severely and negatively impacted. To limit further impacts and partially recover from them, we need to respect measures around social distancing and personal protection.

How can we monitor social distancing and personal protection?

While maintaining social distancing is important to avoid secondary peaks of infections in the short and medium-term, the questions for many public and private organizations are:

How do we monitor and quantify social distancing?

How do we ensure that the measures we take are enough to protect the health of employees and customers?

How do we make sure that preventative resources are allocated in a proactive way where they are needed?

Use existing video infrastructure - make it smart!

Co-Video is a simple yet innovative solution based on image processing with the sole purpose of enabling large scale analysis and auditing of social distancing and use of personal protective measures in common spaces. Our solution will utilize the already existing video security and surveillance infrastructure and securely report an anonymized basic data set which will provide owners of commercial and public spaces the means to audit the extent to which measures are complied with. The solution will be offered as an integrated software and targeted at owners of video security and surveillance infrastructure, ensuring an efficient and impactful large-scale roll-out. Co-Video brings both commercial and societal value by:

Allowing currently closed businesses to more operations more easily: currently many commercial entities (i.e. shopping malls, manufacturing facilities, venues, etc) are unable to resume operation as they are not able to prove that the social distancing and personal protection measures are being complied with. Co-Video will allow for easy auditing of these measures and potentially allow for compliant businesses to resume operations.

Protecting employees and customers from unnecessary risk exposure: real-time imaging analytics through Co-Video would allow for automated risk assessment and alerting where there is continuous or consistent lack of compliance with measures. This will minimize the risk of transmission, protecting employees as well as customers and the general public.

Allowing for targeted efforts and use of public resources: continuous analysis of the data would allow for a broader risk assessment for viral transmission in specific areas within cities such as parks, neighborhoods, markets and others. The proactive identification and prioritization of these spaces would allow for better use of screening measures and enforcement of further public health measures.

What’s in the background?

Co-Video is built around 4 main technological and methodological pillars. After image input Co-Video uses YOLO to detect boundaries and classify objects. The next step involves the measurement of distance which is estimated using triangle similarity technique to the real distance followed by application of the Euclidian length principle to calculate the distance between people. This determines if someone is compliant with social distancing rules. At the same time, the image is cropped using single shot detector and analyzed for the presence of a face mask using a 2D convolutional neural network. The program provides two outputs – a visual one which indicates where the violation of social distancing rules has occurred as well as a numerical one which feeds into an analytical dashboard that provides a global overview. A basic data set export function can allow the sharing of information with third parties through a secure database.

Our method ensures simplicity and does not require further compliance with respect to privacy and sensitive data handing regulations beyond what providers of video surveillance software already need to do.

The market

The European video analytics and video management systems market are estimated to face an approximate growth of 36.87% in the period of 2020-2025. Roughly 60% of this growth will be attributed to the mere outbreak of COVID19 and its disruptive effects. Co-Video targets to reach 3-7% of this growth. We will do so by providing our solution to video surveillance software companies serving infrastructure facilities such as transportation, city surveillance, public places and commercial facilities such as retail stores and malls during and post lockdown.

Our approach is step-wise and vision-based: In a first stage we will focus primarily on the European countries most affected by COVID19 (Spain, Italy). After and in parallel to our first successes in Spain and Italy, we will next target other European countries with a clear market potential for Co-Video. As COVID19 and pandemics in general have long-term consequences, we are convinced that we can remain sustainable thanks to constant monitoring, mapping and adapting of our business activities whilst keeping abreast with latest technological evolutions, conditio sin equa non. (Depending on the specific evolution of COVID19 and in the advent of a vaccine or specific drug therapy we will reevaluate our solution and adapt our Co-Video business model if required.)

Our offer will specifically include customized intelligent visual analytics and visual management solutions. Co-Video uses the latest image processing technologies to move crowd management and face recognition applications to the next level and make them pandemic-adapted. We will provide our clients with the possibility to chose between several software packages through subscription fee, fixed fee, pay per use, specifically adapted to their market segments and needs.

No pain no gain

The development of Co-Video was done entirely as part of the hackathon and with time constraints. As such the team faced and overcame a number of challenges: finding proper data sets (image banks) fast enough to train and test the solution; achieving a more fine-grained object identification whatever the angle µ; refining the distance calculation accounting for focal distance of lens as well as average height per person for each country; alignment on basic the technological approach for the initial prototype; the mechanism for detection of face masks.

The MantuVsVirus team is proud to have developed the Co-Video solution as it shows true bold thinking in action. It is not just a concept from a hackathon - it is a true solution which is simple, innovative, impactful, feasible and marketable!

So what is next?

One of the biggest realizations we have unveiled is that the problems which COVID19 has created are so complex and multidimensional, that only true cross-functional, cross-organizational, cross-industry efforts are required to pull through. It is therefore unlikely that there will be one solution which would fix the damage done but with an appropriate portfolio of solutions the recovery timeline can be improved. For us to succeed in getting through this pandemic efficiently and with minimal long-term consequences, public and private entities need to collaborate closely to ensure broad public safety and eventually economic system integrity.