IDL_Soft_Robot

Project to do object classification using a soft robot finger for 11-785 Intro to IDL.

spiral_finger_data_reader.py Computer-side code to read from serial port and write to .txt file.

IDL_Data_Collection.ino Microcontroller (Arduino) code to run data collection sequence and print to Serial.

environment preparation [plain]

Only need to download pandas, scikit-learn, and torch. You can DIY on your own env, or use the following commands:

conda create -n IDL python=3.8
pip install -r requirements.txt

Training code

• dataloader in utils/dataloader.py
  • read panda dataframes from *.txt files
  • organize data according to columes
  • [TODO] extract features from data
    • currently only return means for each column
• train/eval loop in train.ipynb
  • implement basic MLP for classification based on contact features
  • implement vanilla train loop
  • use scilearn.classification_report to report metrics
    • currently achieve 0.0 validation accuracy :(