SysAg2022

Requirements

Required libraries

pip install opencv-python librosa gdown

Data preparation

Download datasets

python downloadDatasets.py

Splitting

python makeCSV.py

Data Augmentation

The files made by the spectrograms augmentation script are already included in the downloaded Datasets. If you want to make new spectrograms, just delete the "{datasetName}_augmentation" folder in each Dataset folder and run this script.
python spectrograms_aug.py    # --> Make spectrograms
python makeCSV.py             # --> Update CSVs

Training

Start training with custom params

cd ser-experiment/
python train.py --dataset demos --gender male --attention bam --uses_drive True --withAugmentation False --batch_size 64 --epoch 71 --patience 10 etc.

Testing

Start testing with custom params

cd ser-experiment/
python test.py --dataset demos --gender male --attention bam --uses_drive True --batch_size 64 --loadModel ../Results/Models/Gruppo_A/best_model_male-epoch_68.pt
  • Gianmarco Scarano
  • Vincenzo Pio Sgarro
  • Afrim Sokoli