SENTRY

Sentry: Your AI Safe Guard!

What inspires our project, essentially, what is our problem statement?

We focus on making AI more accessible and feasible in the industrial domain by targeting a critical aspect- Ensuring worker safety, especially because the risk of accidents is high. The consequence of an mishap is multi-fold: Would cost more investment on machinery and parts, Halting of work due to failure or repair of machinery or lack of healthy workers, or worse, injuries, diseases, and fatal threats to the workers' lives.

These accidents occur due to two broad causes:

a. Negligence of the workers: simply because they are unaware of the consequences or may be careless.

b. Failure of an machine, or environmental threats, like leakage of a gas tank, or of chemicals. it may also include sharp moving parts, or faulty machinery.

Manually monitoring of workers to check the safety adherence may not be efficient and hence, Our project aims to solve this issue by developing a Real-time Worker Safety Monitoring System that leverages machine learning to detect whether workers are adhering to safety standards. This, of course would be pitched to the companies and would require "admins" to login to the site to access their interactive dashboard.

Our solution uses a machine learning model that can analyze images and detect whether workers are wearing safety gear, such as helmets and life vests. The model processes an input image and returns an annotated image with labels indicating whether the worker is safe or unsafe based on their compliance with safety requirements. This real-time detection system allows supervisors to monitor worker safety remotely and take immediate action when necessary.

•Input: Image or frame of workers on-site.

•Output: Labeled image indicating if the worker is safe (wearing all required gear) or unsafe (missing gear).

This system can reduce human oversight and enhance the overall safety of workers, promoting better compliance with safety regulations in high-risk industries.

Further, upon detection of unsafe behavior by the model,  it would update the components and pages of the dashboard. It shall then alert the main security system as configured by the admin.

We aim to further enhance our model by integrating IoT sensors, which is almost ready to be deployed. We would employ camera and CCTV footages to complement the security. The components of the dashboard would be more dynamic in addition to being responsive, which means, the data detected by the model would directly be updated into the dashboard, making analytics and alerting more quick and efficient. In particular, we are trying to train our model to improve face recognition feature of the unsafe worker. In case of an unsafe behavior detected, the worker's profile section with their status can be tracked and directly reported to the admin who decides how to deal with the worker later.