The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
Error code: JobManagerCrashedError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
image
image |
---|
π Usage with π€ Datasets Library
# ==============================================================================
# Final and reliable method β clean dataset structure, no .cast() required
# ==============================================================================
# Step 1: Install the Hugging Face datasets library
!pip install datasets -q
# Step 2: Download and unzip the dataset (recommended method)
import requests
from zipfile import ZipFile
from io import BytesIO
from pathlib import Path
url = "https://huggingface.co/datasets/makhresearch/skin-lesion-segmentation-classification/resolve/main/skin-lesion-segmentation-classification.zip"
print("Downloading the dataset ZIP file...")
response = requests.get(url)
response.raise_for_status()
print("β
Download complete.")
print("\nExtracting files...")
with ZipFile(BytesIO(response.content)) as zf:
zf.extractall(".")
print("β
β
β
Extraction complete.")
# ==============================================================================
# Done
# ==============================================================================
π§ β¨ Skin Lesion Segmentation & Classification Dataset
Welcome to Skin Lesion Segmentation & Classification β a high-quality medical dataset 𧬠focused on dermatological image analysis using bounding boxes and segmentation annotations. Whether you're training a model for melanoma detection or experimenting with computer vision in healthcare, this dataset is built for you! ππ©Ί
π¦ Dataset Overview
π This dataset follows the YOLO-style folder structure, ready for training in detection or segmentation tasks:
skin-lesion-segmentation-classification/
βββ train/
β βββ images/ # 6,675 training images
β βββ labels/ # YOLO format labels
βββ valid/
β βββ images/ # 1,911 validation images
β βββ labels/
βββ test/
β βββ images/ # 961 test images
β βββ labels/
π Each .txt
label contains YOLO format:
<class_id> x1 y1 x2 y2 x3 y3 ... xn yn
β
All coordinates are normalized between 0
and 1
.
π§ Classes
The dataset supports 7 lesion categories, labeled from ID 0
to 6
. Below is a full list of class codes, descriptions, and emojis for quick visual reference:
π¦ Skin Lesion Types & Description
ID | Emoji | Code | Full Name | Description |
---|---|---|---|---|
0 | π€ | BKL | Benign Keratosis | Non-cancerous, often scaly skin lesions. Common and usually harmless. |
1 | β« | NV | Melanocytic Nevi | Regular moles formed by pigment-producing cells. Typically benign. |
2 | π | DF | Dermatofibroma | Firm, small nodules under the skin caused by minor trauma. Non-cancerous. |
3 | π΄ | MEL | Melanoma | A serious and potentially life-threatening skin cancer. Early detection is critical. |
4 | π΅ | VASC | Vascular Lesion | Blood vessel-related marks like angiomas. Usually red or purple. |
5 | π£ | BCC | Basal Cell Carcinoma | The most common skin cancer. Slow-growing and rarely metastasizes. |
6 | β οΈ | AKIEC | Actinic Keratoses / Intraepithelial Carcinoma | Pre-cancerous lesions that may evolve into squamous cell carcinoma. |
π§ͺ Tasks
- π³ Object Detection (YOLO format with normalized bounding boxes)
- π― Segmentation Approximation (using bounding box areas)
- π Potential for classification/fine-tuning tasks
π¦ Dataset Size
Split | Images | Labels |
---|---|---|
Train | 6,675 | 6,675 |
Valid | 1,911 | 1,911 |
Test | 961 | 961 |
Total | 9,547 | 9,547 |
π Notes
- All lesion images are in
.jpg
format. - Labels are compatible with YOLOv8 training pipelines.
- The dataset is suitable for medical AI, research, and deep learning education.
π§ Use Cases
This dataset can power a wide range of AI tasks:
- π§ͺ Skin lesion detection
- π©» Medical image segmentation
- π€ Disease classification (e.g., melanoma vs benign)
- π Benchmarking object detection models in the medical domain
π·οΈ Classes
π’ Supports one or more lesion classes (e.g., lesion
).
You can edit the class list based on your labeling config.
π License
π Open for academic and commercial use under the MIT License.
π§Ύ Citation
If you use this dataset in your research or project, please cite:
@dataset{makhresearch_skin_lesion_2025,
title = {Skin Lesion Segmentation and Classification Dataset},
author = {MakhResearch},
year = {2025},
url = {https://huggingface.co/datasets/makhresearch/skin-lesion-segmentation-classification}
}
π€ Contributing
π Contributions, feedback, and ideas are always welcome!
Letβs build better medical AI β together. π‘π©Ί
π¬ For questions, citations, or collaboration requests, visit the Hugging Face Profile.
- Downloads last month
- 295