This repository contains the study and report of experiments with the Astro Pi for the European Astro Pi Challenge 2022/23, Mission Space Lab (see details to know some numbers). It is part of the educational repositories to learn how to manage a project.
Before to subcribe the team, we have evaluated some experiments and the public data described here.
What we have learned:
- how to use the tools for this project
- how to read the graphics of the Astro PI metrics
- what the machine learning steps are
What we have learned:
- how to get the Astro PI metrics
- how to recognize the move on the Astro PI metrics graphics
- a simple machine learning algorithm to recognize humans
What we have learned:
- creation of the catalogue of the objects (and people) on the ISS is very challenging
- there are models pre-trained are ready to use
What we have learned:
- PIR sensor detect a person up to approximately 30/15 ft away (details)
- we have used only MotionSensor class
- it would have been interesting to have an ultrasonic distance sensor for using DistanceSensor class
- HOGDescriptor class of opencv-python is less sensitive than the PIR sensor
- humans were not entirely visible in the image
- humans were too close, occupying the whole image with an arm or body