Researchers have developed a new benchmark dataset to address the critical need for verifying the authenticity of satellite imagery, which is increasingly threatened by advanced generative AI and deepfakes. The dataset, containing 60 images with 30 manipulated examples using techniques like copy-paste splicing and diffusion model inpainting, aims to support the training and evaluation of algorithms for detecting and localizing such manipulations. Each image includes ground-truth masks and metadata to facilitate pixel-level analysis and studies on how image collection parameters affect detection performance. AI
IMPACT This dataset will enable better detection of manipulated satellite imagery, crucial for scientific and monitoring applications.
RANK_REASON The cluster describes a new benchmark dataset for research in image forensics and geospatial deepfake detection, published on arXiv.
- copy-paste splicing
- deepfake
- Diffusion Models
- generative artificial intelligence
- geodf/fmow-fake-small
- Hugging Face
- satellite imagery
- arXiv
- Gans
- Philipe Ambrozio Dias
- remote sensing
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
