Adaptive Realtime Detection and Examination Network
An adaptive realtime detection and examination network system,
- Edge deployment
- Visual Classification (dogs, lost children, liscence plate, surfers, etc.)
- Sharpening
- Comparison to uploaded image
- Alert user to located coordinates
The following Table summarizes the scope and topics used
Container | Location | Tx Topic | Rx Topic | Storage | Role |
---|---|---|---|---|---|
Detector (ubuntu) | JetsonNX | image |
N/A |
N/A |
Uses YoloV5 to detect image from Camera and publish to mqtt |
Mqtt broker (alpine) | JetsonNX | image |
image |
N/A |
Pub/Sub local to NX |
Mqtt message forwarder (alpine) | JetsonNX | cloud |
image |
N/A |
Message fowarder, and knowledgeable of multiple MQTT brokers |
Mqtt broker (alpine) | AmazonAWS | cloud |
cloud |
N/A |
Pub/Sub local to AWS EC2 Instance |
Super Resolution (ubuntu) | AmazonAWS | N/A |
cloud |
s3 |
Receives data from NX and enhances image with Super Resolution model |
dji - drone with video use phone wifi hotspot use battery Jetson NX
VisDrone UAV overhead dataset https://github.com/VisDrone/VisDrone-Dataset
DOTA: A Large-scale Dataset for Object Detection in Aerial Images https://captain-whu.github.io/DOTA/index.html
Stanford Drone Dataset http://cvgl.stanford.edu/projects/uav_data/
Other datasets: https://lionbridge.ai/datasets/15-best-aerial-image-datasets-for-machine-learning/