AsmaSima/inception_v3_covid

Inception V3 Covid-19 Classifier

Model description

This model is a fine-tuned Inception V3 convolutional neural network designed for chest X-ray image classification, with a focus on Covid-19 detection.

The model was trained using transfer learning from ImageNet weights and further fine-tuned on medical imaging data.

An optionnal lung auto-masking preprocessing step can be applied upstream to reduce background bias and improve interpretability.

This model is entended for research and educational purposes only.

Intended use

• Binary or multi-class classification of chest X-ray images
• Research on medical image classification
• Demonstration of deep learning pipelines (preprocessing, fine-tuning, Grad-CAM)

Not intended for

• Clinical diagnosis
• Medical decision-making
• Production healthcare systems

Training data

The model was trained on the COVID-19 Radiography Dataset (Kaggle):

https://www.kaggle.com/datasets/tawsifurrahman/covid19-radiography-database/data

The dataset contains chest X-ray images labeled as:

• COVID
• Normal
• Lung-Opacity
• Viral Pneumonia

Only frontal chest X-ray images were considered.

Input format

• Image size : 299x299 px
• Channels : 3 (RBG)
• Pixel values : normalized to [-1,1]

Output format

• Class probabilites (softmax)
• Final prediction via argmax

Training details

• Base model : Inception V3 (ImageNet pretrained)
• Fine-tuning : last 20 layers unfrozen
• Optimizer : SGD
• Loss function : Categorical crossentropy

Evaluation results

The model achieves an accuracy of approximately 84% on the validation set.

Performance metrics were evaluated using :

• Accuracy
• Confusion matrix
• ROC/AUC curves

Results are dataset-dependent and may vary under different data distributions.

Explainability

Model predictions can be interpreted using Grad-CAM visualizations, highlighting salient lung regions contributing to the decision.

Limitations

• Trained on a single public dataset
• Sensitive to image quality and acquisition protocol
• Performance may degrade on non-frontal or low-quality X-ray
• Not validated on clinical-grade datasets

Ethical considerations

This model is provided for research purposes only. It must not be used as a diagnostic tool or to guide clinical decisions.

Example usage

from tensorflow.keras.models import load_model
import numpy as np

model = load_model("inception_V3_covid.keras")

preds = model.predict(input_tensor)

Author

Asma Sima