File size: 4,089 Bytes
0c65f71 a3a4a50 0c65f71 2187224 0c65f71 2187224 0c65f71 2187224 0c65f71 2187224 0c65f71 3a89034 0c65f71 3a89034 0c65f71 3a89034 0c65f71 3a89034 0c65f71 3a89034 0c65f71 3a89034 0c65f71 3a89034 0c65f71 3a89034 0c65f71 3a89034 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 |
---
license: apache-2.0
tags:
- generated_from_trainer
datasets:
- chest-xray-classification
- keremberke/chest-xray-classification
metrics:
- accuracy
pipeline_tag: image-classification
base_model: google/vit-base-patch16-224-in21k
model-index:
- name: vit-xray-pneumonia-classification
results:
- task:
type: image-classification
name: Image Classification
dataset:
name: chest-xray-classification
type: chest-xray-classification
config: full
split: validation
args: full
metrics:
- type: accuracy
value: 0.9742489270386266
name: Accuracy
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# vit-xray-pneumonia-classification
This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on the chest-xray-classification dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0868
- Accuracy: 0.9742
## Inference example
```python
from transformers import pipeline
classifier = pipeline(model="lxyuan/vit-xray-pneumonia-classification")
# image taken from https://www.news-medical.net/health/What-is-Viral-Pneumonia.aspx
classifier("https://d2jx2rerrg6sh3.cloudfront.net/image-handler/ts/20200618040600/ri/650/picture/2020/6/shutterstock_786937069.jpg")
>>>
[{'score': 0.990334689617157, 'label': 'PNEUMONIA'},
{'score': 0.009665317833423615, 'label': 'NORMAL'}]
```
## Training procedure
Notebook link: [here](https://github.com/LxYuan0420/nlp/blob/main/notebooks/ViT-xray-classification.ipynb)
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 15
```python
from transformers import EarlyStoppingCallback
training_args = TrainingArguments(
output_dir="vit-xray-pneumonia-classification",
remove_unused_columns=False,
evaluation_strategy="epoch",
save_strategy="epoch",
logging_strategy="epoch",
learning_rate=5e-5,
per_device_train_batch_size=16,
gradient_accumulation_steps=4,
per_device_eval_batch_size=16,
num_train_epochs=15,
save_total_limit=2,
warmup_ratio=0.1,
load_best_model_at_end=True,
metric_for_best_model="eval_loss",
greater_is_better=False,
fp16=True,
push_to_hub=True,
report_to="tensorboard"
)
early_stopping = EarlyStoppingCallback(early_stopping_patience=3)
trainer = Trainer(
model=model,
args=training_args,
data_collator=data_collator,
train_dataset=train_ds,
eval_dataset=val_ds,
tokenizer=processor,
compute_metrics=compute_metrics,
callbacks=[early_stopping],
)
```
### Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
| 0.5152 | 0.99 | 63 | 0.2507 | 0.9245 |
| 0.2334 | 1.99 | 127 | 0.1766 | 0.9382 |
| 0.1647 | 3.0 | 191 | 0.1218 | 0.9588 |
| 0.144 | 4.0 | 255 | 0.1222 | 0.9502 |
| 0.1348 | 4.99 | 318 | 0.1293 | 0.9571 |
| 0.1276 | 5.99 | 382 | 0.1000 | 0.9665 |
| 0.1175 | 7.0 | 446 | 0.1177 | 0.9502 |
| 0.109 | 8.0 | 510 | 0.1079 | 0.9665 |
| 0.0914 | 8.99 | 573 | 0.0804 | 0.9717 |
| 0.0872 | 9.99 | 637 | 0.0800 | 0.9717 |
| 0.0804 | 11.0 | 701 | 0.0862 | 0.9682 |
| 0.0935 | 12.0 | 765 | 0.0883 | 0.9657 |
| 0.0686 | 12.99 | 828 | 0.0868 | 0.9742 |
### Framework versions
- Transformers 4.30.2
- Pytorch 1.9.0+cu102
- Datasets 2.12.0
- Tokenizers 0.13.3 |