https://github.com/ani717/pneumonia_detection_effecientnet_b7

Pneumonia Detection in Chest X-ray Image with EfficientNet-B7. Accuracy = 87.98%, Precision = 100%, Recall = 83.87%, F1 Score = 91.23.
https://github.com/ani717/pneumonia_detection_effecientnet_b7

cnn computer-vision data-analysis data-augmentation efficientnet image-classification image-processing machine-learning

Last synced: 3 days ago
JSON representation

Pneumonia Detection in Chest X-ray Image with EfficientNet-B7. Accuracy = 87.98%, Precision = 100%, Recall = 83.87%, F1 Score = 91.23.

• Host: GitHub
• URL: https://github.com/ani717/pneumonia_detection_effecientnet_b7
• Owner: ANI717
• License: mit
• Created: 2020-12-12T16:49:39.000Z (almost 4 years ago)
• Default Branch: main
• Last Pushed: 2021-03-25T11:44:57.000Z (over 3 years ago)
• Last Synced: 2023-08-25T11:17:05.321Z (about 1 year ago)
• Topics: cnn, computer-vision, data-analysis, data-augmentation, efficientnet, image-classification, image-processing, machine-learning
• Language: Python
• Homepage:
• Size: 229 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        


  
Pneumonia Detection in Chest X-ray Image with EfficientNet-B7


Google has published both a very exciting paper and source code for a newly designed CNN called EfficientNet, that set new records for both accuracy and computational efficiency over most popular CNNs on ImageNet dataset. Among them, EfficientNet-B7 (which achieved highest accuracy over ImageNet dataset) is used here to solve a classsification problem, detection of Pneumonia in Chest X-ray Images. Without any augmentation, 100% precision can be achieved. 


## Database

Kermany D., Goldbaum M., Cai W. Large dataset of labeled optical coherence tomography (OCT) and chest X-Ray images 2018, 172, 1122–1131. Cell. 2018;172:1122–1131. doi: 10.1016/j.cell.2018.02.010.

https://data.mendeley.com/datasets/rscbjbr9sj/3 


```

Unzip ZhangLabData.zip

Copy it to "./data/" directory

```

Training Dataset:

```

Total Images: 5233

Normal Healthy Person: 1349

Pnumonia Patients: 3884

```

Testing Dataset:

```

Total Images: 624

Normal Healthy Person: 234

Pnumonia Patients: 390

```

Validation Dataset is created from Training Dataset for calibrating Hyperparameters.

## Codebase

Driver Program

```

train.py = runs training session

test.py = runs testing session

folder_to_csv.py = lists files in a folder

merge_csv.py = merges contents in CSV files

augmentation.py = creates augmented dataset

```

Setting File

```

settings.json = contains hyperparameters

```

Utility Classes

```

_datagen.py = data generator for deep learning session

_train_test.py = runs deep learning session

```

EfficientNet by Luke Melas-Kyriazi

https://github.com/lukemelas/EfficientNet-PyTorch

```

./efficientNet/

```

## Result

Plain Data (No Augmentation)

```

Accuracy = 87.98%

Precision = 100%

Recall = 83.87%

F1 Score = 91.23

```