Selective Grasping Using YOLO-GGCNN

Grasping enables the robots to manipulate obstacles or change its state on the workspace. Grasp detection is a basic skill required for robots to perform grasping and manipulation tasks in the workspace. This project demonstrates a two-staged framework where an object is detected and the grasp prediciton is done within that area. The figure below shows the yolo-ggcnn framework.

Installation

This code is tested with Python 3.10.12 on Linux Mint 21.3 and CoppeliaSim 4.6.

Usage

1. Clone this repository.
2. In CoppeliaSim simulator open the scene file yolo-ggcnn.tttlocated in assets folder.
3. Run the run_yolo.py script.
4. Run the run_ggcnn.py script.

Demo

demo.mov

References

1. YOLOv8
2. GG-CNN