krzonkalla/cancer_de_pele_detector

Skin Cancer Image Classification Model

Introduction

Esse modelo classifica imagens de pele em várias categorias, com o objetivo de detectar lesões cancerígenas

Model Overview

• Arquitetura: Vision Transformer (ViT)
• Modelo Pré-treinado: Google's ViT 16x16 treinado no dataset ImageNet21k
• Classification Head Modificada: A classification head foi trocado para adaptar melhor o modelo à nova tarefa

Dataset

• Nome do Dataset: Skin Cancer Dataset HAM10000
• Classes: Benign keratosis-like lesions, Basal cell carcinoma, Actinic keratoses, Vascular lesions, Melanocytic nevi, Melanoma, Dermatofibroma

Training

• Optimizer: Adam optimizer with a learning rate of 1e-4
• Loss Function: Cross-Entropy Loss
• Batch Size: 32
• Number of Epochs: 5

Evaluation Metrics

• Train Loss: Average loss over the training dataset
• Train Accuracy: Accuracy over the training dataset
• Validation Loss: Average loss over the validation dataset
• Validation Accuracy: Accuracy over the validation dataset

Results

• Train Loss: 0.1208
• Train Accuracy: 0.9614
• Val Loss: 0.1000
• Val Accuracy: 0.9695