README.md

metadata

title: Brain Tumor MRI Classifier
emoji: 🧠
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: 5.29.0
app_file: app.py
pinned: false
license: mit
tags:
  - medical-imaging
  - brain-tumor
  - efficientnet
  - image-classification
  - pytorch
  - mri

Brain Tumor MRI Classifier — EfficientNet-B3

A fine-tuned EfficientNet-B3 model for 4-class brain tumor classification from MRI scans, achieving 98.98% validation accuracy and 0.9896 macro F1.

Classes

Class	Description
Glioma	Tumor originating in glial cells of the brain or spine
Meningioma	Tumor arising from the meninges surrounding the brain
Pituitary Tumor	Tumor in the pituitary gland at the base of the brain
No Tumor	No tumor detected in the MRI scan

Model

• Architecture: EfficientNet-B3 (ImageNet pretrained) with custom classification head
• Head: Dropout → Linear(1536, 512) → SiLU → Dropout → Linear(512, 4)
• Input size: 300 × 300
• Training: Two-phase — backbone frozen for 5 epochs (head LR 1e-3), then full fine-tune with differential LR (backbone 1e-4, head 1e-3)
• Schedule: Cosine decay with 3-epoch linear warmup
• Loss: Class-weighted cross-entropy

Weights

The model weights (model.pt) are hosted in this repository and downloaded automatically on first run via huggingface_hub.

To download manually:

from huggingface_hub import hf_hub_download
ckpt_path = hf_hub_download(repo_id="your-hf-username/brain-tumor-efficientnet-b3", filename="model.pt")

Dataset

Trained on a merged dataset from two sources:

• Figshare Brain Tumor Dataset — glioma, meningioma, pituitary MRI scans
• Kaggle Brain Tumor MRI Dataset — 4-class dataset with glioma, meningioma, pituitary, no tumor

Split	Images
Train	8,211
Validation	2,053

Results

Metric	Score
Accuracy	0.9898
Macro F1	0.9896
Weighted F1	0.9898

Per-class F1: Glioma 0.9915 · Meningioma 0.9832 · No Tumor 0.9903 · Pituitary 0.9935

Usage

import torch
import torch.nn as nn
from torchvision import transforms
from torchvision.models import efficientnet_b3
from huggingface_hub import hf_hub_download
from PIL import Image

class EfficientNetClassifier(nn.Module):
    def __init__(self, num_classes=4, dropout=0.4):
        super().__init__()
        self.backbone = efficientnet_b3(weights=None)
        in_features = self.backbone.classifier[1].in_features
        self.backbone.classifier = nn.Sequential(
            nn.Dropout(p=dropout, inplace=True),
            nn.Linear(in_features, 512),
            nn.SiLU(),
            nn.Dropout(p=dropout / 2),
            nn.Linear(512, num_classes),
        )
    def forward(self, x):
        return self.backbone(x)

# Load
ckpt_path = hf_hub_download("your-hf-username/brain-tumor-efficientnet-b3", "model.pt")
ckpt = torch.load(ckpt_path, map_location="cpu", weights_only=False)
id_to_label = {int(k): v for k, v in ckpt["id_to_label"].items()}

model = EfficientNetClassifier()
model.load_state_dict(ckpt["model"])
model.eval()

# Infer
transform = transforms.Compose([
    transforms.Resize((300, 300)),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
])

img = Image.open("mri_scan.jpg").convert("RGB")
with torch.no_grad():
    probs = torch.softmax(model(transform(img).unsqueeze(0)), dim=-1)[0]
    pred  = id_to_label[probs.argmax().item()]

print(pred)

Disclaimer

This model is intended for research purposes only and is not a certified medical diagnostic tool. Do not use for clinical decision-making.