Just look up zoom nudity incident, and you’ll see the many embarrassing moments recorded on video conferencing apps for the world to see. People accidentally leaving their cameras on as they get dressed, or family members revealed undressing or nude in the background.
These accidents can have terrible consequences. Exposing oneself on camera, even accidentally, violates the sexual harassment policies of companies and universities and has lead to countless terminated employment contracts, scholarships, etc. Our goal at this hackathon was to help ensure the safety, privacy and dignity of the millions of people who use webcams worldwide.
What it does
Camblocker is a platform that aims to bring camera safety and privacy solutions to the masses. We use machine learning models to ensure that no embarrassing or exposing footage is captured on camera.
We recognize that different users will have different needs and concerns so we created multiple solutions- the user can access our changing and nudity detection models from a chrome extension, a python script, or a hardware prototype.
How we built it
Changing Detection: The model built in the teachable machine using our own data in different scenarios and angles. Chrome Extension: Used JS to close tabs if it detects clothing removal AppSafe: Python script which quits an app if clothing removal is detected. CamBlocker Pro Hardware Solution: SG90 mini servo motor, an esp32 microcontroller (Arduino core), and various 3d printed parts and screws, a total of around 20 dollars.
Challenges we ran into
Since there were so many parts to our project, the integration of all parts was one of the main challenges that we ran into during the hackathon. The main challenge was the integration of the chrome extension to the DCP backend. One project where we encountered the most challenge was the DCP implementation. It was because we had not encountered DCP before the hackathon so it required us to adjust our environments and learn the documentation in DCP.
Accomplishments that we're proud of
We are proud that we were able to provide multiple different solutions to the problem with a limited timeframe. We are also proud that we were able to learn new tools like teachable machines and TensorFlow during the process and have fun building important programs that we were passionate about.
What we learned
We also learned to use a teachable machine and TensorFlow to do machine learning on custom datasets. We also learned about python to serial connection in our hardware implementation.
What's next for Cam Blocker
We hope that by providing a variety of options for our users, we can encourage them to adopt our detection models. We also think we can expand this to CCTV and home camera systems. That way, even if a person's camera footage gets hacked, they have the peace of mind that the attacker won’t have any footage of themselves or their family.