|
--- |
|
license: apache-2.0 |
|
base_model: google/vit-base-patch16-224-in21k |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- imagefolder |
|
model-index: |
|
- name: ViT-Breast-Cancer |
|
results: [] |
|
widget: |
|
- src: https://pathology.jhu.edu/build/assets/breast/_gallery/invasive-lobular-carcinoma.jpg |
|
example_title: Invasive Lobular Carcinoma |
|
|
|
pipeline_tag: image-classification |
|
--- |
|
|
|
# ViT-Breast-Cancer |
|
|
|
This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on a dataset of breast cancer microscope slides. |
|
|
|
## Model description |
|
|
|
This is a fine-tuned ViT (Google) that serves more as an exploration of vision transformers in medicine for my learning than as anything specific. I fine-tuned this model on a dataset of ~7000 images of breast cancer slides labelled as 'benign' or 'cancerous'. |
|
I used the Transformers library and the out-of-the-box ViTForImageClassification configuration. |
|
Despite this being an incredibly barebones fine-tune, I hope you fine it useful! Any recommendations are welcome! |
|
|
|
## Intended uses & limitations |
|
|
|
This is a super basic fine tuned model. Please evaluate its performance for yourself do determine whether it can be useful for you. In a big picture sense, this model can tell apart benign and cancerous breast tissue samples. |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 0.0002 |
|
- train_batch_size: 16 |
|
- eval_batch_size: 8 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
|
|
### Training results |
|
|
|
- training_loss = 0.007100 |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.42.0.dev0 |
|
- Pytorch 2.3.0+cu121 |
|
- Datasets 2.20.0 |
|
- Tokenizers 0.19.1 |
