<h1>deit-base-distilled-patch16-224-Brain_Tumors_Image_Classification</h1>
This model is a fine-tuned version of facebook/deit-base-distilled-patch16-224.
It achieves the following results on the evaluation set:
- Loss: 1.8587
- Accuracy: 0.8046
- Weighted f1: 0.7749
- Micro f1: 0.8046
- Macro f1: 0.7814
- Weighted recall: 0.8046
- Micro recall: 0.8046
- Macro recall: 0.7996
- Weighted precision: 0.8567
- Micro precision: 0.8046
- Macro precision: 0.8710
<div style="text-align: center;"> <h2> Model Description </h2> <a href="https://github.com/DunnBC22/Vision_Audio_and_Multimodal_Projects/blob/main/Computer%20Vision/Image%20Classification/Multiclass%20Classification/Brain%20Tumors%20Image%20Classification%20Comparison/DeiT%20-%20Image%20Classification.ipynb"> Click here for the code that I used to create this model </a> This project is part of a comparison of seventeen (17) transformers. <a href="https://github.com/DunnBC22/Vision_Audio_and_Multimodal_Projects/blob/main/Computer%20Vision/Image%20Classification/Multiclass%20Classification/Brain%20Tumors%20Image%20Classification%20Comparison/README.md"> Click here to see the README markdown file for the full project </a> <h2> Intended Uses & Limitations </h2> This model is intended to demonstrate my ability to solve a complex problem using technology.
<h2> Training & Evaluation Data </h2> <a href="https://www.kaggle.com/datasets/sartajbhuvaji/brain-tumor-classification-mri"> Brain Tumor Image Classification Dataset </a> <h2> Sample Images </h2> <img src="https://github.com/DunnBC22/Vision_Audio_and_Multimodal_Projects/raw/main/Computer%20Vision/Image%20Classification/Multiclass%20Classification/Brain%20Tumors%20Image%20Classification%20Comparison/Images/Sample%20Images.png" /> <h2> Class Distribution of Training Dataset </h2> <img src="https://github.com/DunnBC22/Vision_Audio_and_Multimodal_Projects/raw/main/Computer%20Vision/Image%20Classification/Multiclass%20Classification/Brain%20Tumors%20Image%20Classification%20Comparison/Images/Class%20Distribution%20-%20Training%20Dataset.png"/> <h2> Class Distribution of Evaluation Dataset </h2> <img src="https://github.com/DunnBC22/Vision_Audio_and_Multimodal_Projects/raw/main/Computer%20Vision/Image%20Classification/Multiclass%20Classification/Brain%20Tumors%20Image%20Classification%20Comparison/Images/Class%20Distribution%20-%20Testing%20Dataset.png"/> </div>
Training procedure
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
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Weighted f1 | Micro f1 | Macro f1 | Weighted recall | Micro recall | Macro recall | Weighted precision | Micro precision | Macro precision |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1.6561 | 1.0 | 180 | 1.5974 | 0.7792 | 0.7454 | 0.7792 | 0.7524 | 0.7792 | 0.7792 | 0.7722 | 0.8318 | 0.7792 | 0.8488 |
| 1.6561 | 2.0 | 360 | 1.7614 | 0.7944 | 0.7575 | 0.7944 | 0.7633 | 0.7944 | 0.7944 | 0.7896 | 0.8458 | 0.7944 | 0.8582 |
| 0.172 | 3.0 | 540 | 1.8587 | 0.8046 | 0.7749 | 0.8046 | 0.7814 | 0.8046 | 0.8046 | 0.7996 | 0.8567 | 0.8046 | 0.8710 |
Framework versions
- Transformers 4.28.1
- Pytorch 2.0.0
- Datasets 2.11.0
- Tokenizers 0.13.3
