Skip to content

Naif-W-Alharthi/Physical-and-Digital-Attacks-on-Grasping-Networks

Repository files navigation

Adversial attack on grasping network

Package Overview

A package to preform digital and phyiscal attacks using SPEA2 on a grasping network1.

Installation

$ git clone https://github.com/Naif-W-Alharthi/Phyiscal-and-Digital-Attacks-on-Grasping-Networks.git

Change directories into the repository and run the pip installation.
$ pip install .

Arguments

Variable Description
-method Takes a string that should be: driver or physical and will determine the type of simluation used.
-m Takes a string that selectes the model used can take All which will simluate the attacks using all models in /models. Can take Lowest which will simluate using the model with the least attempts, otherwise it will use the model provided according to model look up table
-allfigs Takes a string that should be True or False. Creates all the diagarm.
-barplot Takes a string that should be True or False. Creates the bar plot of both lowest quality and radius/Intensity change in the models.
-Mu Takes an int. Mu of the simluation.
-Lambda Takes an int. Lambda of the simluation
-Ngen Takes an int. Ngen of the simluation
-Min_Stra Takes an int. MIN_STRATEGY of the simluation
-Max_Stra Takes an int. MAX_STRATEGY of the simluation

Note: if a value is not mentioned from the ones needed to run a simluation they will be replaced with the default value.

Model look up table

String Model used for simluaion
"gqcnn_suction" GQCNN-4.0-SUCTION
fc-gqcnn_suction FC-GQCNN-4.0-SUCTION
fc-gqcnn_parellel_grasp FC-GQCNN-4.0-PJ
dex2.1 Dexnet2.1
dex2.0 Dexnet2.0

tested with python 3.9 and above.

Footnotes

  1. [Reference to grasping networks].

About

Python module to preform attacks on grasping network and create diagrams.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published