RepVGG-GELAN is a novel deep-learning model designed for accurate and efficient brain tumour detection in medical images. It integrates RepVGG, a reparameterized convolutional approach, into the YOLO framework, enhancing speed and precision. The model also leverages the Generalized Efficient Layer Aggregation Network (GELAN) to further improve feature extraction and detection performance.

• Enhanced YOLO architecture by incorporating RepVGG for optimized object detection.
• Efficient feature extraction by utilizing GELAN for improved accuracy and speed.
• State-of-the-art performance by Achieving superior results compared to existing RCS-YOLO models.
• Optimized for medical imaging, tailored for brain tumor detection using deep learning.

The model is trained and evaluated on the Brain Tumour Detection 2020 (Br35H) dataset from Kaggle. This dataset consists of 701 medical images labeled for tumour detection.

To set up the environment and run the model, follow these steps:

# Clone the repository
git clone https://github.com/ThensiB/RepVGG-GELAN.git
cd RepVGG-GELAN

# Install dependencies
pip install -r requirements.txt

The model is implemented using PyTorch and trained on Google Colab with an NVIDIA RTX 3090 GPU.

# Train the model
python train.py --data data.yaml --epochs 150 --batch-size 8 --img-size 640

RepVGG-GELAN combines:

• RepVGG Blocks: Reparametrized convolutional networks for efficient learning.
• GELAN Blocks: Spatial Pyramid Pooling and multi-scale feature extraction.
• ADown Modules: Asymmetric downsampling for better localization.

The model's performance is measured using:

• Precision & Recall: Measures the accuracy of predictions.
• mAP (mean Average Precision): Evaluates model accuracy over different IoU thresholds.
• FLOPs (Floating Point Operations): Assesses computational efficiency.

Run the following command to perform inference on test images:

python detect.py --weights best.pt --source test_images/

If you use this work, please cite:

@article{balakrishnan2024repvgg-gelan,
  title={RepVGG-GELAN: Enhanced GELAN with VGG-STYLE ConvNets for Brain Tumour Detection},
  author={Balakrishnan, Thennarasi and Sengar, Sandeep Singh},
  journal={arXiv preprint arXiv:2405.03541},
  year={2024}
}

This research was conducted at Cardiff Metropolitan University. Special thanks to the contributor and the open-source community for their valuable resources. Anyone is free to use this code.