prajval10 / Supervised-Autoencoder-Joint-Learning-on-Heterogeneous-Tactile-Sensory-Data

Supervised-Autoencoder-Joint-Learning-on-Heterogeneous-Tactile-Sensory-Data

This is the supplementary material for the paper Supervised Autoencoder Joint Learning on Heterogeneous Tactile Sensory Data: Improving Material Classification Performance, International Conference on Intelligent Robots and Systems (IROS) 2020.

Requirements

To run the code, you can use the following command to install required packages in a seperate virtual environment.

conda create -n AE python=3.8.3
conda activate AE
pip install -r requirements.txt

Datasets

In this experiment, we use data of two tactile sensors, iCub and BioTac.
gh_download.sh contains terminal commands to download the preprocessed data from a separate git repo BioTac_slide_20_50
usage: open gh_download, change DIR to a preferred location, then run

chmod 755 gh_download.sh 
./gh_download.sh 

After saving the data, you can begin to train a model by simply running the following command under code folder python T1_BT19_Icub_joint_ae.py -k 0 -c 0 --data_dir /home/ruihan/data.
More examples and explanations for parameters are given at the beginning of individual code script.

Folders and files

• code contains all the code used to run the experiments and analyze the results.
  • models_and_stats is an empty folder to contained trained models and saved statistics
• docs contains supplementary information (empty for now).

Hope you find it helpful and please feel free to raise issues and questions. We are always happy to help and learn from each other! (^ω^)