Grasping Model with Enhanced SE(2) Prediction

Project Overview

The Grasping Model with Enhanced SE(2) Prediction project aimed to improve robotic grasping capabilities through the development of a PyTorch-based model. By incorporating safe-z prediction into the SE(2) transformation model, the project optimized the model for patch-based input, enhancing its performance in robotic grasping tasks.

Key Features

• Safe-Z Prediction: Integrated safe-z prediction into the SE(2) model to enhance the accuracy of grasping height predictions, ensuring safer and more reliable robotic grasping.
• Efficiency Improvements: Reduced the training process to 300 epochs, achieving a doubling in speed compared to the 600-epoch baseline, without compromising the model’s performance.