gianlab's picture
Update README.md
5961a88
metadata
license: apache-2.0
base_model: microsoft/swin-tiny-patch4-window7-224
tags:
  - generated_from_trainer
datasets:
  - imagefolder
metrics:
  - accuracy
model-index:
  - name: swin-tiny-patch4-window7-224-finetuned-parkinson-classification
    results:
      - task:
          name: Image Classification
          type: image-classification
        dataset:
          name: imagefolder
          type: imagefolder
          config: default
          split: train
          args: default
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.9090909090909091

swin-tiny-patch4-window7-224-finetuned-parkinson-classification

This model is a fine-tuned version of microsoft/swin-tiny-patch4-window7-224 on the imagefolder dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4966
  • Accuracy: 0.9091

Model description

This model was created by importing the dataset of spiral drawings made by both parkinsons patients and healthy people into Google Colab from kaggle here: https://www.kaggle.com/datasets/kmader/parkinsons-drawings/data. I then used the image classification tutorial here: https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/image_classification.ipynb

obtaining the following notebook:

https://colab.research.google.com/drive/1oRjwgHjmaQYRU1qf-TTV7cg1qMZXgMaO?usp=sharing

The possible classified data are:

  • Healthy
  • Parkinson

Spiral drawing example:

Screenshot

Intended uses & limitations

Acknowledgements

The data came from the paper: Zham P, Kumar DK, Dabnichki P, Poosapadi Arjunan S and Raghav S (2017) Distinguishing Different Stages of Parkinson’s Disease Using Composite Index of Speed and Pen-Pressure of Sketching a Spiral. Front. Neurol. 8:435. doi: 10.3389/fneur.2017.00435

https://www.frontiersin.org/articles/10.3389/fneur.2017.00435/full

Data licence : https://creativecommons.org/licenses/by-nc-nd/4.0/

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-05
  • train_batch_size: 32
  • eval_batch_size: 32
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 20

Training results

Training Loss Epoch Step Validation Loss Accuracy
No log 1.0 1 0.6801 0.4545
No log 2.0 3 0.8005 0.3636
No log 3.0 5 0.6325 0.6364
No log 4.0 6 0.5494 0.8182
No log 5.0 7 0.5214 0.8182
No log 6.0 9 0.5735 0.7273
0.3063 7.0 11 0.4966 0.9091
0.3063 8.0 12 0.4557 0.9091
0.3063 9.0 13 0.4444 0.9091
0.3063 10.0 15 0.6226 0.6364
0.3063 11.0 17 0.8224 0.4545
0.3063 12.0 18 0.8127 0.4545
0.3063 13.0 19 0.7868 0.4545
0.2277 14.0 21 0.8195 0.4545
0.2277 15.0 23 0.7499 0.4545
0.2277 16.0 24 0.7022 0.5455
0.2277 17.0 25 0.6755 0.5455
0.2277 18.0 27 0.6277 0.6364
0.2277 19.0 29 0.5820 0.6364
0.1867 20.0 30 0.5784 0.6364

Framework versions

  • Transformers 4.35.2
  • Pytorch 2.1.0+cu121
  • Datasets 2.16.1
  • Tokenizers 0.15.0