Deep-Learning-based-automated-trash-sorting-system

automated trash sorting system using YOLOV8x
can categrize 4 trash type

• can, paper, glass_bottle, pet

Test Env.

The code is tested successfully at

• Python 3.8.10
• Linux 22.04 LTS
• YOLOv8
  • model: yolov8x.pt

Hardware Requirements

• OpenMANIPULATOR-X (RM-X52-TNM)
• U2D2 motor control board
• RealSense D435
• 3d printed D435 holder

Requirements

YOLOv8 Setting

• Install YOLOv8 on a machine by ultralytics git link
  can download our pre-trained model from google drive

Dynamixel Setting

• Install pip install dynamixel-sdk package by ROBOTIS

pip install dynamixel-sdk

Python Setting

• To run the python code, following pakages are necessary
• to solve forward/inverse kinematics, we used ikpy

pip install ikpy
pip install pyrealsense2	
pip install opencv-python

Model data

• transfer learned 250 ~ 600 pics on each categroy based
  aihub - 생활 폐기물 이미지

How to Run

| Before run, set all the motor velocity and acceleration by DYNAMIXEL Wizard to 70

• for gipper(motor ID = 15), use default speed and accel

main.py : main code to run
teleop.py : teleoperation arm

Results