ModelPrésentéExterne

Skin Lesion Ensemble

Sourcehuggingface.co/kindai-derma-ai/skin-lesion-ensemble/tree/05a3e6c8e6a994154589f80673d2e2b368ab743e

TypeOtherVersionEfficientnet b4 ham10k fold0CClasse

Source du modèle

Type: Other
Taille du fichier: 67,7–332 MB
Version: vEfficientnet b4 ham10k fold0
Mis à jour: 10 oct. 2026
Source: Hugging Face

Description de la source

7-Model Ensemble for Skin Lesion Classification

8クラス皮膚病変分類のための7モデルアンサンブルシステム Test Accuracy: 95.92%

Model Description

35個のディープラーニングモデル(7種類 × 5-fold CV)を組み合わせた高精度な皮膚病変診断支援システムです。

Target Classes (8 classes)

Class	Japanese	English
ADM	後天性真皮メラノサイトーシス	Acquired Dermal Melanocytosis
BCC	基底細胞癌	Basal Cell Carcinoma
Ephelis	雀卵斑	Freckles
Melanoma	悪性黒色腫	Malignant Melanoma
Melasma	肝斑	Melasma
Nevus	色素性母斑	Melanocytic Nevus
SK	脂漏性角化症	Seborrheic Keratosis
Solar Lentigo	日光黒子	Solar Lentigo

Available Model Configurations

このリポジトリには2つのモデル構成が含まれています:

Configuration	Models	Folder	Use Case
Base Models	20 models	`base_models/`	シンプルなアンサンブル、軽量版
Full Ensemble	35 models	Root folders	最高精度、バイナリ精緻化あり

Base Models (20 models) - バイナリなし版

base_models/
├── efficientnet_b4_ham10k/      # EfficientNet-B4 × 5
├── efficientnet_b7_imagenet/    # EfficientNet-B7 × 5
└── transformers/                 # ViT × 5, Swin × 5

Full Ensemble (35 models) - フル版

├── efficientnet_b4_ham10k/      # EfficientNet-B4 × 5
├── efficientnet_b7_imagenet/    # EfficientNet-B7 × 5
├── transformers/                 # ViT × 5, Swin × 5
└── binary_classifiers/          # Binary classifiers × 15

Performance

Metric	Value
Test Accuracy	95.92%
Macro Precision	96.09%
Macro Recall	95.56%
Macro F1-Score	95.77%
Melanoma Sensitivity	100%
Melanoma Specificity	100%

Per-Class Performance

Class	Precision	Recall	F1-Score
ADM	96.30%	96.30%	96.30%
Basal Cell Carcinoma	87.50%	91.30%	89.36%
Ephelis	100.00%	94.44%	97.14%
Malignant Melanoma	95.24%	100.00%	97.56%
Melasma	97.14%	100.00%	98.55%
Nevus	100.00%	94.44%	97.14%
Seborrheic Keratosis	95.65%	88.00%	91.67%
Solar Lentigo	96.88%	100.00%	98.41%

Model Architecture

7 Model Types (35 total models)

Model	Pre-training	Parameters	Count
EfficientNet-B4	ImageNet → HAM10000	~19M	× 5 folds
EfficientNet-B7	ImageNet	~66M	× 5 folds
Vision Transformer (ViT)	ImageNet	~86M	× 5 folds
Swin Transformer	ImageNet	~88M	× 5 folds
Binary: ADM vs Nevus	-	~19M	× 5 folds
Binary: Nevus vs Melanoma	-	~19M	× 5 folds
Binary: Solar vs Ephelis	-	~19M	× 5 folds

Ensemble Strategy

Stage 1: Base Ensemble (20 models)

Average softmax probabilities from all base models

Stage 2: Binary Refinement (15 models) - Optional

Apply binary classifiers to refine confusing pairs:
- ADM ↔ Nevus
- Nevus ↔ Melanoma
- Solar Lentigo ↔ Ephelis

Usage

Installation

pip install torch torchvision timm pillow huggingface_hub

Download Models

from huggingface_hub import snapshot_download

# Option 1: Download base models only (20 models, ~5GB)
snapshot_download(
    repo_id="kindai-derma-ai/skin-lesion-ensemble",
    local_dir="./models",
    allow_patterns="base_models/**"
)

# Option 2: Download full ensemble (35 models, ~6.4GB)
snapshot_download(
    repo_id="kindai-derma-ai/skin-lesion-ensemble",
    local_dir="./models",
    ignore_patterns="base_models/**"  # Avoid duplicates
)

# Option 3: Download everything
snapshot_download(
    repo_id="kindai-derma-ai/skin-lesion-ensemble",
    local_dir="./models"
)

Inference with Base Models

import torch
import timm
from PIL import Image
from torchvision import transforms

# Load model (example with EfficientNet-B4)
model = timm.create_model('efficientnet_b4', pretrained=False, num_classes=8)
checkpoint = torch.load('models/base_models/efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold0.pth')
model.load_state_dict(checkpoint['model_state_dict'])
model.eval()

# Preprocessing
transform = transforms.Compose([
    transforms.Resize((384, 384)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])

# Predict
image = Image.open('skin_lesion.jpg').convert('RGB')
input_tensor = transform(image).unsqueeze(0)

with torch.no_grad():
    output = model(input_tensor)
    probabilities = torch.softmax(output, dim=1)
    predicted_class = torch.argmax(probabilities, dim=1).item()

class_names = ['ADM', 'BCC', 'Ephelis', 'Melanoma', 'Melasma', 'Nevus', 'SK', 'Solar_Lentigo']
print(f"Prediction: {class_names[predicted_class]}")
print(f"Confidence: {probabilities[0][predicted_class]:.2%}")

File Structure

models/
├── base_models/                      # Base models only (20 models)
│   ├── efficientnet_b4_ham10k/
│   ├── efficientnet_b7_imagenet/
│   └── transformers/
│
├── efficientnet_b4_ham10k/           # Full ensemble components
├── efficientnet_b7_imagenet/
├── transformers/
└── binary_classifiers/               # Binary refinement (full ensemble only)

Training Details

Input Size: 384 × 384
Loss Function: Focal Loss (α=1, γ=2)
Optimizer: AdamW (lr=1e-4, weight_decay=0.01)
Scheduler: CosineAnnealingLR
Batch Size: 16
Epochs: 50 (with early stopping, patience=15)
Cross-Validation: 5-fold stratified

Data Augmentation

RandomHorizontalFlip
RandomVerticalFlip
RandomRotation (±20°)
ColorJitter (brightness, contrast, saturation, hue)
RandomAffine (translation, scale)

Limitations and Intended Use

Intended Use

Research and educational purposes
Dermatology AI research
Clinical decision support (with physician oversight)

Limitations

Not for clinical diagnosis: This model is intended as a decision support tool and should not replace professional medical diagnosis
Dataset bias: Trained primarily on Japanese patient data; performance may vary on other populations
Image quality dependency: Performance depends on image quality and standardized acquisition conditions

Out-of-Scope Use

Autonomous clinical diagnosis without physician review
Use on non-dermatoscopic images
Commercial deployment without proper validation

Citation

@misc{skin_lesion_ensemble_2024,
  title={7-Model Ensemble for 8-Class Skin Lesion Classification},
  author={Kindai Dermatology AI Lab},
  year={2024},
  howpublished={\url{https://huggingface.co/kindai-derma-ai/skin-lesion-ensemble}}
}

License

This model is released under CC BY-NC 4.0.

Academic/Research Use: Freely available
Commercial Use: Contact authors for licensing

Acknowledgments

Created: December 2024 Organization: Kindai Dermatology AI Lab

Sources

1 source

Hugging FacePrincipalVersion 05a3e6c8e6a994154589f80673d2e2b368ab743e

huggingface.co/kindai-derma-ai/skin-lesion-ensemble/tree/05a3e6c8e6a994154589f80673d2e2b368ab743e

Vérifié 10 oct.

Download files

Artefacts du modèle

55 artefacts

base_models/efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold0.pth

pth · 202 MB · SHA-256 62b5f5dac63e…f727 · Hugging Face

Download file

base_models/efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold1.pth

pth · 202 MB · SHA-256 0d5850306b56…a565 · Hugging Face

Download file

base_models/efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold2.pth

pth · 202 MB · SHA-256 4705b7762111…8886 · Hugging Face

Download file

base_models/efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold3.pth

pth · 202 MB · SHA-256 02ce6ff2b3cf…90cc · Hugging Face

Download file

base_models/efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold4.pth

pth · 202 MB · SHA-256 af972d4e35ba…a345 · Hugging Face

Download file

base_models/efficientnet_b7_imagenet/improved_efficientnet_b7_fold0.pth

pth · 245 MB · SHA-256 26868190a73f…7204 · Hugging Face

Download file

base_models/efficientnet_b7_imagenet/improved_efficientnet_b7_fold1.pth

pth · 245 MB · SHA-256 0b46947bcac7…7e66 · Hugging Face

Download file

base_models/efficientnet_b7_imagenet/improved_efficientnet_b7_fold2.pth

pth · 245 MB · SHA-256 10ae01e8c75f…3b2a · Hugging Face

Download file

base_models/efficientnet_b7_imagenet/improved_efficientnet_b7_fold3.pth

pth · 245 MB · SHA-256 90dd7ffb63d3…359b · Hugging Face

Download file

base_models/efficientnet_b7_imagenet/improved_efficientnet_b7_fold4.pth

pth · 245 MB · SHA-256 eaabe9c0f940…477c · Hugging Face

Download file

base_models/transformers/swin_transformer_fold0.pth

pth · 332 MB · SHA-256 2d46243fb253…b4b6 · Hugging Face

Download file

base_models/transformers/swin_transformer_fold1.pth

pth · 332 MB · SHA-256 1457cba88fd0…fc87 · Hugging Face

Download file

base_models/transformers/swin_transformer_fold2.pth

pth · 332 MB · SHA-256 d7e6672ed29e…6e6b · Hugging Face

Download file

base_models/transformers/swin_transformer_fold3.pth

pth · 332 MB · SHA-256 33ea68931288…5bf7 · Hugging Face

Download file

base_models/transformers/swin_transformer_fold4.pth

pth · 332 MB · SHA-256 eb5614b62c4c…55ac · Hugging Face

Download file

base_models/transformers/vit_transformer_fold0.pth

pth · 328 MB · SHA-256 c35d3ad3da14…b58b · Hugging Face

Download file

base_models/transformers/vit_transformer_fold1.pth

pth · 328 MB · SHA-256 68e4a3821566…b33d · Hugging Face

Download file

base_models/transformers/vit_transformer_fold2.pth

pth · 328 MB · SHA-256 8a8e8ac66bd9…7e29 · Hugging Face

Download file

base_models/transformers/vit_transformer_fold3.pth

pth · 328 MB · SHA-256 7b7cc0a55203…6fcc · Hugging Face

Download file

base_models/transformers/vit_transformer_fold4.pth

pth · 328 MB · SHA-256 267ccaef0c59…6fd4 · Hugging Face

Download file

binary_classifiers/binary_adm_vs_nevus_fold0.pth

pth · 67,7 MB · SHA-256 1e3f2204841a…0f0e · Hugging Face

Download file

binary_classifiers/binary_adm_vs_nevus_fold1.pth

pth · 67,7 MB · SHA-256 98025e177e65…4989 · Hugging Face

Download file

binary_classifiers/binary_adm_vs_nevus_fold2.pth

pth · 67,7 MB · SHA-256 c5e46599d693…1377 · Hugging Face

Download file

binary_classifiers/binary_adm_vs_nevus_fold3.pth

pth · 67,7 MB · SHA-256 f0d28d443162…5054 · Hugging Face

Download file

binary_classifiers/binary_adm_vs_nevus_fold4.pth

pth · 67,7 MB · SHA-256 134836b72c84…f82d · Hugging Face

Download file

binary_classifiers/binary_nevus_vs_melanoma_fold0.pth

pth · 67,7 MB · SHA-256 806fe90ef23f…3b00 · Hugging Face

Download file

binary_classifiers/binary_nevus_vs_melanoma_fold1.pth

pth · 67,7 MB · SHA-256 f016eebe94db…dccd · Hugging Face

Download file

binary_classifiers/binary_nevus_vs_melanoma_fold2.pth

pth · 67,7 MB · SHA-256 10fd17866b42…8383 · Hugging Face

Download file

binary_classifiers/binary_nevus_vs_melanoma_fold3.pth

pth · 67,7 MB · SHA-256 d7ef347f2750…a40d · Hugging Face

Download file

binary_classifiers/binary_nevus_vs_melanoma_fold4.pth

pth · 67,7 MB · SHA-256 6859282bbd37…269e · Hugging Face

Download file

binary_classifiers/binary_solar_vs_ephelis_fold0.pth

pth · 67,7 MB · SHA-256 e37447bcc988…d7f0 · Hugging Face

Download file

binary_classifiers/binary_solar_vs_ephelis_fold1.pth

pth · 67,7 MB · SHA-256 42575b5acb77…26c3 · Hugging Face

Download file

binary_classifiers/binary_solar_vs_ephelis_fold2.pth

pth · 67,7 MB · SHA-256 f9c614a3da85…680a · Hugging Face

Download file

binary_classifiers/binary_solar_vs_ephelis_fold3.pth

pth · 67,7 MB · SHA-256 050e0bf2b845…44fd · Hugging Face

Download file

binary_classifiers/binary_solar_vs_ephelis_fold4.pth

pth · 67,7 MB · SHA-256 b533174016d0…d1b0 · Hugging Face

Download file

efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold0.pth

pth · 202 MB · SHA-256 62b5f5dac63e…f727 · Hugging Face

Download file

efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold1.pth

pth · 202 MB · SHA-256 0d5850306b56…a565 · Hugging Face

Download file

efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold2.pth

pth · 202 MB · SHA-256 4705b7762111…8886 · Hugging Face

Download file

efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold3.pth

pth · 202 MB · SHA-256 02ce6ff2b3cf…90cc · Hugging Face

Download file

efficientnet_b4_ham10k/efficientnet_b4_ham10k_fold4.pth

pth · 202 MB · SHA-256 af972d4e35ba…a345 · Hugging Face

Download file

efficientnet_b7_imagenet/improved_efficientnet_b7_fold0.pth

pth · 245 MB · SHA-256 26868190a73f…7204 · Hugging Face

Download file

efficientnet_b7_imagenet/improved_efficientnet_b7_fold1.pth

pth · 245 MB · SHA-256 0b46947bcac7…7e66 · Hugging Face

Download file

efficientnet_b7_imagenet/improved_efficientnet_b7_fold2.pth

pth · 245 MB · SHA-256 10ae01e8c75f…3b2a · Hugging Face

Download file

efficientnet_b7_imagenet/improved_efficientnet_b7_fold3.pth

pth · 245 MB · SHA-256 90dd7ffb63d3…359b · Hugging Face

Download file

efficientnet_b7_imagenet/improved_efficientnet_b7_fold4.pth

pth · 245 MB · SHA-256 eaabe9c0f940…477c · Hugging Face

Download file

transformers/swin_transformer_fold0.pth

pth · 332 MB · SHA-256 2d46243fb253…b4b6 · Hugging Face

Download file

transformers/swin_transformer_fold1.pth

pth · 332 MB · SHA-256 1457cba88fd0…fc87 · Hugging Face

Download file

transformers/swin_transformer_fold2.pth

pth · 332 MB · SHA-256 d7e6672ed29e…6e6b · Hugging Face

Download file

transformers/swin_transformer_fold3.pth

pth · 332 MB · SHA-256 33ea68931288…5bf7 · Hugging Face

Download file

transformers/swin_transformer_fold4.pth

pth · 332 MB · SHA-256 eb5614b62c4c…55ac · Hugging Face

Download file

transformers/vit_transformer_fold0.pth

pth · 328 MB · SHA-256 c35d3ad3da14…b58b · Hugging Face

Download file

transformers/vit_transformer_fold1.pth

pth · 328 MB · SHA-256 68e4a3821566…b33d · Hugging Face

Download file

transformers/vit_transformer_fold2.pth

pth · 328 MB · SHA-256 8a8e8ac66bd9…7e29 · Hugging Face

Download file

transformers/vit_transformer_fold3.pth

pth · 328 MB · SHA-256 7b7cc0a55203…6fcc · Hugging Face

Download file

transformers/vit_transformer_fold4.pth

pth · 328 MB · SHA-256 267ccaef0c59…6fd4 · Hugging Face

Download file

Extraits de sources

2 extraits

Readme1 aoûthuggingface.co/kindai-derma-ai/skin-lesion-ensemble

--- license: cc-by-nc-4.0 language: - ja - en tags: - image-classification - medical - dermatology - skin-lesion - ensemble - pytorch - efficientnet - vision-transformer - swin-transformer datasets: - HAM10000 pipeline_tag: image-classification metrics: - accuracy - f1 - precision - recall library_name: timm --- # 7-Model Ensemble for Skin Lesion Classification 8クラス皮膚病変分類のための7モデルアンサンブルシステム **Test Accuracy: 95.92%** ## Model Description 35個のディープラーニングモデル（7種類 × 5-fold CV）を組み合わせた高精度な皮膚病変診断支援システムです。 ### Target Classes (8 classes) | Class | Japanese | English | |-------|----------|---------| | ADM | 後天性真皮メラノサイトーシス | Acquired Dermal Melanocytosis | | BCC | 基底細胞癌 | Basal Cell Carcinoma | | Ephelis | 雀卵斑 | Freckles | | Melanoma | 悪性黒色腫 | Malignant Melanoma | | Melasma | 肝斑 | Melasma | | Nevus | 色素性母斑 | Melanocytic Nevus | | SK | 脂漏性角化症 | Seborrheic Keratosis | | Solar Lentigo | 日光黒子 | Solar Lentigo | ## Available Model Configurations このリポジトリには2つのモデル構成が含まれています： | Configuration | Models | Folder | Use Case | |---------------|--------|--------|----------| | **Base Models** | 20 models | `base_models/` | シンプルなアンサンブル、軽量版 | | **Full Ensemble** | 35 models | Root folders | 最高精度、バイナリ精緻化あり | ### Base Models (20 models) - バイナリなし版 ``` base_models/ ├── efficientnet_b4_ham10k/ # EfficientNet-B4 × 5 ├── efficientnet_b7_imagenet/ # EfficientNet-B7 × 5 └── transformers/ # ViT × 5, Swin × 5 ``` ### Full Ensemble (35 models) - フル版 ``` ├── efficientnet_b4_ham10k/ # EfficientNet-B4 × 5 ├── efficientnet_b7_imagenet/ # EfficientNet-B7 × 5 ├── transformers/ # ViT × 5, Swin × 5 └── binary_classifiers/ # Binary classifiers × 15 ``` ## Performance | Metric | Value | |--------|-------| | **Test Accuracy** | **95.92%** | | Macro Precision | 96.09% | | Macro Recall | 95.56% | | Macro F1-Score | 95.77% | | **Melanoma Sensitivity** | **100%** | | **Melanoma Specificity** | **100%** | ### Per-Class Performance | Class | Precision | Recall | F1-Score | |-------|-----------|--------|----------| | ADM | 96.30% | 96.30% | 96.30% | | Basal Cell Carcinoma | 87.50% | 91.30% | 89.36% | | Ephelis | **100.00%** | 94.44% | 97.14% | | Malignant Melanoma | 95.24% | **100.00%** | 97.56% | | Melasma | 97.14% | **100.00%** | 98.55% | | Nevus | **100.00%** | 94.44% | 97.14% | | Seborrheic Keratosis | 95.65% | 88.00% | 91.67% | | Solar Lentigo | 96.88% | **100.00%** | 98.41% | ## Model Architect...

Note1 aoûthuggingface.co/kindai-derma-ai/skin-lesion-ensemble

Source context: 0 downloads · 0 likes · Pipeline image-classification · Library timm · Repo kindai-derma-ai/skin-lesion-ensemble