LesNet (Lesion Net) is an open-source project for AI-based skin lesion detection. It aims to create a reliable tool and foster community involvement in critical AI problems. Contributions are welcome!
Hyperparameter Tuning: Integrated hyperparameter tuning for the SkinCancerDetector model using Keras Tuner, aiming to optimize model performance on the skin cancer dataset.
Data Scraper Improvements: Updated the data scraper for robustness and efficiency. It now gracefully handles restarts, avoiding re-downloads of existing files, and iterates through all available pages of the dataset.
Model Adjustments: Modified the model architecture to better handle binary classification of skin lesions into benign and malignant categories. Adjustments include layer modifications and the integration of dropout layers to reduce overfitting.
Error Handling: Enhanced error handling mechanisms during model training and data processing to ensure smoother execution and debugging.
Purpose
These updates aim to enhance the model's accuracy and reliability in classifying skin cancer images, while also ensuring efficient and effective data collection from the ISIC Archive. By improving our model's training process and data handling, we can push forward in our mission to develop a more accurate skin cancer detection tool.
Key Updates
SkinCancerDetectormodel using Keras Tuner, aiming to optimize model performance on the skin cancer dataset.Purpose
These updates aim to enhance the model's accuracy and reliability in classifying skin cancer images, while also ensuring efficient and effective data collection from the ISIC Archive. By improving our model's training process and data handling, we can push forward in our mission to develop a more accurate skin cancer detection tool.