Kaggle ISIC2024 3rd place solution

3rd place solution for ISIC 2024 - Skin Cancer Detection Challenge

Environment

• WSL2 Ubuntu 20.04.6 LTS
• GPU RTX3090 x 1

Setup python environment

python -m venv .venv
source ./.venv/bin/activate
pip install -r requirements.txt

Data

Place the competiion data into input directory

cd ./input
kaggle competitions download -c isic-2024-challenge
unzip isic-2024-challenge.zip -d isic-2024-challenge
cd ../

Prepare fold

python define_fold.py

Train image models

model1

convnextv2_nano.fcmae_ft_in22k_in1k (CV 0.1599389)

python train_20240827234748.py

Model weights are save to ./output_20240827234748

model2

vit_tiny_patch16_224.augreg_in21k_ft_in1k (CV 0.1612504)

python train_20240830205516.py

Model weights are save to ./output_20240830205516

model3

vit_tiny_patch16_224.augreg_in21k_ft_in1k (CV 0.1460650)

python train_20240831025049.py

Model weights are save to ./output_20240831025049

model4

vit_small_patch16_224.augreg_in21k_ft_in1k (CV 0.1486832)

python train_20240902001446.py

Model weights are save to ./output_20240902001446

Trained models are uploaded in kaggle dataset. 20240827234748 20240830205516 20240831025049 20240902001446

Integrate image into tabular model and submission

tabular-with-image-submission.ipynb