top of page

Pipeline

Our data pipeline consists of three primary pieces an edge, device, the messages client, and cloud infastructure

The Edge

  • The edge device is triggered once a new object of interest enters the camera's field of view. 

  • The event is logged and sent to the MQTT message broker

Our edge software is containerized and device-independent and has been shown to work on numerous edge devices like raspberry pi, Nvidea Jetson, Ring Doorbell, and more ...

edge.png

Message Broker

mqtt_edited.jpg

We chose Mosquitto MQTT as our messaging broker between the edge device and the cloud. MQTT is the IOT industry standard because of its ease of use, lightweight footprint and high quality of service

cloud3.png

The Cloud

Our cloud infrastructure is composed of five docker containers. These containers receive video messages from MQTT then process the video that has been sent to the cloud and store's it in AWS S3 for future viewing on our website www.argosiswatching.com. 

​

The video processing steps include

  • Object recognition using Yolov5

  • A sequence-to-sequence NLP model for creating a video caption

  • A custom lookup feature base on the objects identified in the video

  • Facial recognition for visitors that have been seen before

  • Daily email service containing video short list of daily events

bottom of page