Hello there! And welcome to the Melanoma W&B project. This project contains all experiments and other W&B features that I utilized to get in the top-5% of the Kaggle ISIC Melanoma Classification competition.

If you want to replicate all tables, charts, sweeps and embedding projection visualization - follow the instructions in the Melanoma W&B GitHub repository.

If you haven't already seen the reports, feel free to check them out here - reports. There are 4 reports that are part of this project and below I provide a small introduction to each of those:

• How to prepare the dataset for the Melanoma Classification?: This report showcases how to download and prepare the dataset for the Melanoma competition. We also store the dataset similar to the Weights and Biases tables as below called "Train data" and "Validation data".
• How to track all your experiments using Microsoft Excel?: In this report, I showcase how I used W&B to track all experiments that I ran as part of the Melanoma competition. We learn how to build beautiful dashboards such as the one below as part of this report.
• How to save all your trained model weights locally after every epoch: As part of this report I showcase how I utilized W&B artifacts to store all model weights for each of the experiments after every epoch. This makes it really easy for me later when ensembling models to pick the best performing models and create an ensemble for top-5 or top-10 models.
• How to Build a Robust Medical Model Using Weights & Biases: As part of this report, I showcase how I integrated various products of Weights and Biases ecosystem such as Tables, Sweeps, Artifacts and Embedding Projector to really get in the top-5% of the leaderboard on Kaggle!

If you have any questions about this project - please feel free to reach out to me at aman@wandb.com.