Brain Tumour Classification

Recently, finished a small machine learning project to classify different types of brain cancer from MRI images.

Project Overview:

• Goal: Classify different types of brain tumours from MRI scans.
• Model: A four layer CNN with max pooling, Normalisation, ReLU activation functions, Dropout, and trained over 5 epochs.
• Tools: Python, PyTorch

Data and Features:

The dataset is loaded using torchvision.datasets.ImageFolder and torch.utils.data.DataLoader. The following transofrmation are applied

• Resize: Images are resized to 224x224 pixels.
• Random Horizontal Flip: Applied to training data for augmentation
• ToTensor: Converts image to PyTorch tensors.
• Normalisation: Uses mean [0.485, 0.456, 0.406] and standard deviation [0.229, 0.224, 0.225]

Training and Results:

• Training Setup: I trained the model with the Adam optimiser (learning rate: 0.001) and CrossEntropyLoss as the loss function.
• Weight Decay: 1e-4 (0.0001)
• Train Loss: 0.407
• Train Accuracy: 85.45%
• Test Loss: 0.7143
• Test Accuracy: 75.97%

Available here and image below.