LuxeFats/Brain-Tumor-Detection-FasterRCNN

Faster R-CNN Object Detection Application

• —An end-to-end object detection system for detecting brain tumors in MRI images using Faster R-CNN with ResNet-50 backbone.
• —The system allows users to upload images via a web interface and receive detected objects with bounding boxes in real time.
• —As of January 7, 2026, the application is a web-based object detection system built using PyTorch, FastAPI, and React.

Results

The following results demonstrate the performance of the Faster R-CNN model on validation MRI images after 10 training epochs.

• —mAP@0.5: 0.69 [image]
• —mAP@0.5-0.9: 0.49 [image]

Problem Statement

• —Brain tumor detection in MRI images is a critical task in medical imaging.
• —Manual annotation is time-consuming and prone to human error.
• —This project aims to automatically detect tumor regions using deep learning–based object detection.

Architecture

Browser (Frontend)
    |
    | HTTP (Image Upload)
    v
Backend API (FastAPI)
    |
    | Load Model & Run Inference
    v
Faster R-CNN (PyTorch)
    |
    v
Detection Results (Annotated Image)

Dataset

• —Source: Kaggle – Brain-tumor dataset by Ultralytics
• —Documentation: https://docs.ultralytics.com/datasets/detect/brain-tumor/
• —Total images: 1116
• —Train: 893
• —Test: 223
• —Dataset configuration: brain-tumor.yaml
• —Classes: Brain tumor
• —Annotation format: YOLO

Methodology

1. 1.Data preprocessing
2. 2.Train Faster R-CNN with ResNet-50
3. 3.Loss optimization (classification + bounding box regression)
4. 4.Evaluation using Precision, Recall, mAP
5. 5.Deployment-ready inference accessible via localhost web interface

Model & Training

• —Framework: PyTorch (Torch and TorchVision)
• —Backbone: ResNet-50
• —Optimizer: SGD
• —Learning rate: 0.005
• —Epochs: 10
• —Hardware: NVIDIA GPU T4

Project Structure

├── backend/                # FastAPI
│   └── main.py
├── checkpoints/
│   └── model.pth
├── data/
│   └── raw/
├── frontend/               # React / HTML / CSS
│   ├── src/
│   │   ├── App.css
│   │   ├── App.tsx
│   │   ├── index.css
│   │   └── main.tsx
│   ├── .gitignore
│   ├── index.html
│   ├── package.json
│   └── vite.config.ts
├── models/
│   ├── faster_rcnn_scratch.ipynb
│   └── train.ipynb
├── .gitignore
├── README.md
└── requirements.txt

Installation

1. Clone this repo

git clone https://github.com/NguyenHuuPhat2203/Web-Based-Faster-R-CNN-Object-Detection-System.git
cd Web-Based-Faster-R-CNN-Object-Detection-System
python -m venv .venv
source .venv/bin/activate  # On Windows use `.venv\Scripts\activate`
pip install -r requirements.txt

2. Training (Deep Learning)

• —Open models/train.ipynb in Jupyter or VS Code or google colab (preferred).
• —Run cells to train the model and save the checkpoint to checkpoints/model.pth or manually download from google colab or google drive.

3. Backend (FastAPI)

• —Run the server:

cd backend
uvicorn main:app --reload

• —The API will be available at http://localhost:8000.

4. Frontend (React)

• —Install dependencies:

cd frontend
npm install

• —Run development server:

npm run dev

• —Open the provided URL:http://localhost:8001.

Future Works

• —Try YOLO models for faster inference
• —Deploy the model as a portable application
• —Real-time detection using webcam

Author(s)

Nguyen Huu Phat Email: phat.nguyenluxefats@gmail.com GitHub: https://github.com/NguyenHuuPhat2203