2 classic AI on edge IoT camera connectivity solutions

Edge intelligent camera is an important internet of things (IoT) device that senses and understands the world, structures video streams to event alarms with edge AI.

If we use OpenNCC edge AI camera to complete artificial intelligence (AI) inference, which is powered by Intel VPU and OpenVINO, we need to convert the extracted AI results into the event storage of interest. At this time, the camera needs to become a part of the IoT. We could integrate Raspberry Pi with OpenNCC to build an AIOT IPC, there are generally two AIOT connectivity solutions:

• Cloud-based IoT
• Peer-to-peer (P2P) based IoT

Cloud-based IoT

We connect AIoT cameras to cloud services so that cloud services can manage devices, store data, and process data uniformly.

The advantages of this scheme are:

• There are a large number of mature solution providers to facilitate quick device access
• Unified data management in the cloud provides natural advantages for continuous model training and iteration of AI services.

The disadvantage is that:

• Data stored in the cloud has privacy risks
• Cloud storage and management have ongoing expenses, and business models that cannot be continuously charged face great challenges

AWS IoT Platform

We shared how to build an Edge AI camera using Raspberry PI and OpenNCC, but using the same solution could build AIOT IPC. We have a Turnkey edge AI IPC, you could integrate it into your own IoT platform.

Ali IoT Platform

We’ve customized similar software solutions for our clients before, and we’ve open-sourced the basics. In this case, the Ali IoT Platform is used. The camera runs the face detection model, When a face is detected, MQTT Is used to send a message to the cloud platform. The python demo can be found at the OpenNCC Hub Repo.

Connect AIoT camera using Nabto’s Peer-to-peer (P2P)

A P2P (peer to peer) connection is a direct communication infrastructure between two peers: a client device (such as a smartphone or a laptop) and an Edge AI-IOT device (such as a surveillance camera, smart door lock, alarm system, heat controller, or anything else that can connect to the internet).

P2P solutions also require Cloud servers, compared to all services in the Cloud(AWS IoT). P2P servers are used to help the terminal establish an encrypted communication link with the IoT device and clients. Once the communication is established, the cloud service does not need to be involved in the subsequent sensitive data interaction. We could find that P2P solutions do not store sensitive data (such as photos) in the cloud, images can be stored on the client device or IoT camera side.

Eyecloud.ai used Nabto’s P2P to develop a Sticker-Eye Camera, one of the earliest battery-powered AI-on-Edge cameras. Now we can also deploy Nabto’s solution on our OpenNCC IP4 to build a P2P, high-security, intelligent camera.

If you have purchased our OpenNCC camera and want to make an edge ai P2P powered IP camera with Raspberry PI, you can apply for Nabto authorization ID from our FAE.