The team and inspiration :
We are a group of 4 passionate data scientists and developers wanting to contribute to create a better comeback from the covid pandemic and boost up the economy again. We are blessed by being mentored by amazing people and industry leaders giving us realistic feedback on the feasibility of the solution.
Our project comes as a direct solution to a very real and alarming problem within manufacturing businesses. Currently, all factories are stopped in Canada and most countries, and should they start again, would need a preventive solution so they can operate in a productive COVID-free environment.
What is the problem :
Due to the high virality of the covid-19 and governmental protocols, factories have been shut down along with most businesses. However, that was just a temporary emergency solution, and soon enough they should open again. This worries both the government and the medical crew, as there haven’t been strict protocols yet to ensure a covid-safe environment for the workers.
The best way actually to minimize the spreading of the pandemic is practicing social distancing and wearing preventive gear in forms of masks and gloves. This could be a challenge to implement in factories where the environment encourages proximity, and tracking the respect of these procedures can prove to be a challenge.
That’s where our solution comes in place, enabling an in-depth monitoring of the safety levels of the factories. This would allow rapid intervention in the event of a critical violation of these protocols.
How our solution works :
General concept :
The solution is an all-in-one platform that allows in-depth monitoring of factories ensuring that workers respect a certain distance between each other while wearing covid-preventive gear (medical masks) and also measuring their body temperature when entering and leaving the factory to check for an eventual fever should the temperature stay high for a certain amount of time.
An operator would be monitoring a dashboard receiving real time feed of information and alerts concerning different units and plants. The factory can be broken down into different units, either by spatial occupancy or process. Then each unit would have a score, raising each time a violation of covid security protocols occurs (while also appearing as an alert on the dashboard).
In the current state of the project and of covid-security protocol, we identified three metrics to monitor : Respect of a minimal social distance Workers must wear face masks permanently inside the factory. Employees with high body temperature hinting to an eventual fever should not be allowed to enter the facility
Respect of minimal social distance :
We do this by using RFID chips embedded in employees access cards, that should allow to estimate each employee’s position and thus track the relative distance between all employees at all time. A minimal distance to respect must be set, and every time it is violated, a notification is pushed to the monitoring platform and the unit’s risk score raises.
Workers must wear face masks :
Cameras backed with a state of the art AI algorithm would allow to check how many people are not wearing masks, and that number would add to the respective unit’s risk score.
Employees having a high temperature would be asked to stay at home as it can be a symptom, and is thus dangerous.
Checking the temperature of every employee that enters the facility using a thermometer would prove difficult. So we thought of a way to measure the temperature of employees without wasting their time, or creating lines and bottlenecks at the door of the factory. In that line of thought, we will be implementing thermal cameras that would automatically measure their temperature and report the high-risk cases to the monitoring platform.
The monitoring platform :
All information regarding the safety of the factory and its units will be consolidated into a platform. A real time feed of alerts and statistics concerning each unit are available. That allows both to monitor the risk and security level of each unit covid-wise and for a fast intervention should a critical situation arise.
Employee’s security and privacy :
This whole project has as its core the security of the employees. Indeed, each employee should come to work with a reassurance that he is working in a safe environment, and it’s the employer’s duty to ensure that. This project is a tool to make this claim possible.
To ensure employee’s privacy, they will not be scored individually, but as a unit. The score in itself is not aimed to penalize them, but is a way to assess the risk factor of each unit and the factory as a whole.
How we built it :
NodeJS & Mongodb stack to build the platform’s logic used for real time monitoring.
Pretrained state of the art mask recognition AI algorithm that can be further improved with data collected insite.
A simulated factory site with Python and Celluloid Package. A set of Workers are generated and start performing a random walk in the factory site with respect to geometrical constraints. Positions are broadcasted to the monitoring platform and notifications are generated in the system.
Future work :
Apply the pretrained model on another context distribution while levergaring the model performances using false positives to train the model in an online learning perspective.
Connect different parts of the platform as it has multiple functionalities.
Extend the app to other environments.
Challenges faced :
Reassuring the plant workers.
What are we proud of :
Being able to build everything from scratch in less than 48 hours.
Having a wonderful and diversified team, experts in their domain.