Dataset · 1aurent
100k histological images of intestinal tissue
1aurent/NCT-CRC-HE
A dataset of 100,000 histological image patches of colorectal cancer and normal tissue.
What's inside
Contains 100,000 non-overlapping histological image patches of human tissues (colorectal cancer and normal) stained with hematoxylin and eosin. All images are 224x224 pixels (0.5 microns per pixel) and color-normalized using Macenko's method. The dataset is designed for medical image classification.
What you can do with it
- Training classifiers to detect pathologies in histological sections
- Pre-training or adapting vision models for digital pathology
- Testing biomedical image analysis algorithms
- Tasks
- Image classification
- Size
- 100K–1M rows
- Downloads this month
- 23.3K
- Likes
- 6
Models trained on it
- 1aurent/vit_small_patch8_224.lunit_dinoImage classification 2.1K
- 1aurent/resnet50.tiatoolbox-kather100kImage classification 1.3K
- 1aurent/vit_small_patch16_224.lunit_dinoImage classification 1.2K
- 1aurent/resnet18.tiatoolbox-kather100kImage classification 942
- 1aurent/resnet50.lunit_btImage classification 623
- 1aurent/densenet121.tiatoolbox-kather100kImage classification 575
- 1aurent/resnet50.lunit_swavImage classification 417
- 1aurent/resnet34.tiatoolbox-kather100kImage classification 393
How to load it
from datasets import load_dataset
ds = load_dataset("1aurent/NCT-CRC-HE")