Year 5 Project
Data
Copernicus Database
Applications
Hardware
Current
Cargo/Pirate tracking
Herd Tracking
Wild Fire Tracking
Satalite
Raspberry PI
msp430
16-bit microcontroller
top speed 25 MHz
Low power consumption
Future?
Odroid board?
Discontiued
Client
HARDWARE CHANGE Wants new chip
Report
Stats
Execution Times
Pink Board?
FPGA?
IS A ACCELLERATOR?
Very little power consumption
For Machine learning on the edge is this federated learning?
Used for data centers, to be more sustainable(in what way gussing power consuption?)
Alba Orbital Data
Result
Christmas Goal
study on low power accelerated machine learning on satellites
Built ML Model
Come to the University with the satellite, 3rd to 7th Oct
Literature Review
Accelerated Machine Learning
Pico Satallite
Lens
Camera
POWER SUPPLY
Radio
How to do this
Offloading tasks to other satellites which are not currently using camera to look at images
General questions
What is inference in this context?
Compare ML approaches
Power Consumption
Only send images that are useful back
Different Hardware (boards?)
CPU vs GPU?
Machine Learning on the EDGE
HARDWARE???
See some examples of the training Data
What software is used for this low power accelerated ML?
Can i make a classifier which sends back anything that looks diffrent? Update to current img when needed
LOW POWER HARDWARE
Will this replace msp430?
Accuracy