Dataset · danielaivanova
ARTeFACT benchmark for analog media defect segmentation
danielaivanova/damaged-media
A collection of images and masks for evaluating analog media damage segmentation models.
What's inside
The ARTeFACT dataset contains up to a thousand examples (images and text metadata) featuring various types of analog media degradation. The card includes ready-to-use PyTorch code for loading data, splitting by material or content type, and visualizing examples.
What you can do with it
- Testing semantic segmentation models for defects and artifacts
- Training neural networks for automated restoration of vintage photos and film
- Classifying damage types on digitized media
- Tasks
- Image → image, Image segmentation, Image → text, Image classification
- Size
- up to 1K rows
- Downloads this month
- 700
- Likes
- 9
How to load it
from datasets import load_dataset
ds = load_dataset("danielaivanova/damaged-media")