Moraeus Vision

Welcome to my computer vision page where curiosity meets too much free time.

Tracking robot

This Frankenstein-looking robot follows the person who raises their hand. Showing the hand again tells the robot to stop following and look for another person.The robot runs on a Nvidia Jetson Orin Nano and is built with ROS2, using YOLO for person detection and MediaPipe for hand gesture recognition.

Bird watcher

This app runs on a Raspberry Pi and live streams to a website. If the user chooses to classify the bird, a detection model first looks for birds in the image, and a second classification model identifies the species and provides some facts about it.An egg detection model checks for eggs once an hour and notifies subscribers by email if any are found.

Counting objects in production

This app lets you draw the line over a video frame and each lime crossing the line is counted.The model is a fine-tuned RF-DETR model.

Pose detection

This app is just to showcase what can be done with pose detection. Each limb has several points which can be used to track movements.Here we use Mediapipe for pose detection.

Fall detection

Fall detection can be used to automatically identify when a person has fallen, enabling rapid response in environments such as nursing homes or railway tracks.Could be optimized with Mediapipe.

Drone detection

Simple drone tracking made from a fine-tuned RF-DETR model.