You can deploy this model and start earning money today!

mdsajjadullah/
chest-xray-pneumonia-resnet50

Views

No views yet

Chest X-Ray Pneumonia Detection — ResNet-50

Author: Md. Sajjad Ullah

Model Description

Fine-tuned ResNet-50 for binary chest X-ray classification (NORMAL vs PNEUMONIA). Trained on 5,216 images with weighted sampling to handle class imbalance (1:2.9 ratio).

Results (Test Set — 624 images)

Metric	Score
Accuracy	91.83%
ROC-AUC	96.50%
Recall (Pneumonia)	98.21%
F1-Score	93.76%

Training Details

Base model : ResNet-50 (ImageNet pretrained)
Fine-tuned : Layer3, Layer4, FC
Optimizer : AdamW (lr=1e-4, weight_decay=1e-4)
Scheduler : CosineAnnealingLR
Augmentation : RandomCrop, Flip, Rotation, ColorJitter
Regularization: Label smoothing 0.1 + Dropout 0.4

How to Use

python

import torch
from torchvision.models import resnet50

CLASS_NAMES = ['NORMAL', 'PNEUMONIA']

model = resnet50()
model.fc = torch.nn.Sequential(
    torch.nn.Dropout(0.4),
    torch.nn.Linear(2048, 2)
)
model.load_state_dict(
    torch.load("best_ResNet-50_.pth", map_location="cpu")
)
model.eval()

Dataset

Chest X-Ray Images (Pneumonia) — Paul Mooney / Kaggle

Kaggle Dataset

https://www.kaggle.com/datasets/mdsajjadullah/chest-x-ray-pneumonia-classification

mdsajjadullah/chest-xray-pneumonia-resnet50