Privacy Preserving Smart Cameras

Inspiration

I started Ambianic.ai while looking for privacy preserving smart cameras than can help me with home automation. Quickly realized that most commercial products have compromised privacy due to storing camera footage in the cloud.

What it does

Ambianic.ai is an open source project that implements on-device inference such as fall detection, person detection, face identification and others. Then allows users to connect directly to their edge devices over WebRTC and see a timeline of detection events. Users can also choose to receive push notifications.

How we built it

Using open source building blocks, Vue, Vuetify, PeerJS on the client side and Raspberry Pi, Python and Tensorflow Lite for the Edge Device/ Smart Camera.

Challenges we ran into

P2P frameworks are generally less mature, harder to work with and harder to test than client/server ones.

Accomplishments that we're proud of

The project has released several stable versions and we continue to make improvements. In 2020 we received several awards, including Google AI Tensorflow Project Spotlight award.

What we learned

User Data Privacy is a well known and sensitive topic that most big tech companies are not ready to solve.

What's next for Privacy Preserving Smart Cameras

We will work on adding more AI models on-device such as license plate identification via federated learning, face identification and others. We will also add more notification channels.

Built With

• javascript
• python
• tensorflow
• vue
• vuetify