sns-homogenization-forensics-dataset

# Dataset DescriptionThis dataset contains the experimental data supporting the research paper “Device-Aware SNS Processing with Confidence Calibration: Breaking the Detection-Confidence Tradeoff in Image Forensics ## Directory Structure & Dataset Distribution### Device Statistics (Device_001 to Device_009)| Device ID | Device Name | Device Type | Dataset Set | Total Images ||-----------|----------------------|-------------|-------------------|--------------|| device_001 | Xiaomi Redmi | Smartphone | Training | 419 || device_002 | Honor V9 | Smartphone | Training | 417 || device_003 | Honor V20 | Smartphone | Calibration | 501 || device_004 | Honor V200 | Smartphone | Test (Phone) | 354 || device_005 | Apple iPhone | Smartphone | Calibration | 300 || device_006 | Samsung Galaxy | Smartphone | Test (Phone) | 300 || device_007 | Motorola | Smartphone | Test (Phone) | 300 || device_008 | Sony NEX-7 | Camera | Test (Camera) | 300 || device_009 | Additional Device | Smartphone | Additional | 213 || Total | 9 Devices | -- | -- | 3104 | ### View Condition Distribution- view_000 (Normal/Unprocessed): ~1,040 images (33.5%)- view_001 (QQ-processed): ~1,040 images (33.5%)- view_002 (Sina Weibo-processed): ~1,024 images (33.0%) ### Dataset Partition by Usage- **Training Set**: 836 images (26.9%) - device_001 (Xiaomi Redmi): 419 images - device_002 (Honor V9): 417 images- **Calibration Set**: 801 images (25.8%) - device_003 (Honor V20): 501 images - device_005 (Apple iPhone): 300 images- **Test Set (Phone)**: 954 images (30.7%) - device_004 (Honor V200): 354 images - device_006 (Samsung Galaxy): 300 images - device_007 (Motorola): 300 images- **Test Set (Camera)**: 300 images (9.7%) - device_008 (Sony NEX-7): 300 images– **Additional Set**: 213 images (6,9 %) - device_009: 213 images ## Usage Instructions1. This dataset includes images under three processing conditions: original images, QQ-processed images und Sina-Weibo-processed images.2. The dataset is clearly partitioned by device type and usage scenario, facilitating model training and calibration.3. It is recommended to use the dataset in accordance with the experimental design specified in the associated research paper. ## Citation FormatIf you use this dataset in your research, please cite the relevant research paper. ## LicenseThis dataset is licensed under [please select a license, e.g. CC BY 4.0]. ## Contact InformationFor questions or inquiries about the dataset, please contact:**Ningxian Zhu**Corresponding AuthorSchool of Cyberspace SecurityShandong University of Political Science and LawJinan, 250014, ChinaEmail: [e-mail]