What it does
With no cure of COVID-19 in sight for at least another year or so, our best weapon against it is to prevent it. Safety Insight is an AI-powered surveillance system which detects and alerts officials if people aren't wearing protective gear - masks, gloves, coats, etc
How I built it
- Step 1 - Was to run the model on sagemaker and understand it's input/output requirements
- Step 2 - Deciding the use cases
- Step 3 - Building a minimal frontend which captures the camera input, sends a screenshot to the backend to be analyzed
- Step 4 - Calling the endpoint and returning the results to frontend
- Step 5 - Using AWS Polly to alert the user .
- Step 6 - Integrated Sendgrid's mail API to notify the user on hazard alert
Challenges I ran into
Running sagemaker model endpoints locally
Running the ML models on sagemaker and creating an endpoint was straightforward. But to get it working locally was a bit frustrating. The key was to understand AWS's IAM thoroughly - users, roles, permissions, zones.
Accomplishments that I'm proud of
- Building an entire application from end-to-end!
- Integrating a pre-trained ML model.
What I learned
- Getting started with cloud
- Understanding AWS Sagemaker concepts
What's next for Safety Insights
- Automated periodic checks
- Allow multiple video sources