IllustrationMOOSE
Aerial person detection. Computer vision on a fixed-wing UAV, connected to ATAK.
Explore the projectRobotics / Embedded systems
A small car. A complete perception loop.
Traffic-sign recognition, a Raspberry Pi and an ultrasonic sensor bring perception into motion.
01 / The challenge
The question behind the system.
Connect a trained image classifier with the timing and physical constraints of a moving platform.
An educational robotic car links a camera-equipped Raspberry Pi to a Flask inference server. A CNN classifies stop, left and right signs, and the client translates the returned decision into motor and steering commands. Ultrasonic sensing adds obstacle detection.
02 / The approach
From architecture to implementation.
- 01
Train a CNN to distinguish the three traffic-sign classes.
- 02
Capture images on the Raspberry Pi, send them to a Flask API and apply the returned movement command.
- 03
Read ultrasonic distance and integrate an obstacle-triggered stop with motor and steering control.
03 / Outcomes
What the work shows.
- The repository brings together model training, inference serving, camera capture and embedded vehicle control.
04 / Scope & next steps
Where the work stands.
- Lane detection, behavioural cloning and measured telemetry are listed as future work. This is an educational platform, not a road-driving system.
Field notes / Visual documentation
A closer look.
Continue exploring
