nebula/CDDB

1---2pretty_name: CDDB3task_categories:4- image-classification5task_ids:6- multi-class-image-classification7tags:8- deepfake9- continual-learning10- computer-vision11- image-forensics12- wacv13size_categories:14- unknown15annotations_creators:16- no-annotation17language:18- en19license: unknown20configs:21- config_name: default22  data_files:23  - split: train24    path: CDDB.tar25dataset_info:26  features: []27---28 29# Dataset Card for CDDB30 31## Dataset Description32 33CDDB is a benchmark dataset introduced in the WACV 2023 paper *A Continual Deepfake Detection Benchmark: Dataset, Methods, and Essentials*.34It is designed for continual deepfake detection, where manipulated images from different deepfake generation sources arrive sequentially instead of being observed all at once.35 36The benchmark is intended to evaluate both:37 38- binary deepfake detection (real vs. fake)39- continual and incremental learning under distribution shifts across deepfake sources40 41Compared with conventional static deepfake datasets, CDDB focuses on a more realistic setting in which new manipulation methods appear over time and a detector must adapt without catastrophically forgetting previously seen sources.42 43## Supported Tasks44 45- Binary image classification: real vs. fake46- Multi-source deepfake classification47- Continual learning / class-incremental learning48- Domain generalization and robustness evaluation for deepfake detection49 50## Dataset Sources51 52- Paper: [A Continual Deepfake Detection Benchmark: Dataset, Methods, and Essentials](https://openaccess.thecvf.com/content/WACV2023/html/Li_A_Continual_Deepfake_Detection_Benchmark_Dataset_Methods_and_Essentials_WACV_2023_paper.html)53- Code repository: [Coral79/CDDB](https://github.com/Coral79/CDDB)54 55## Paper Information56 57**Title:** A Continual Deepfake Detection Benchmark: Dataset, Methods, and Essentials58 59**Authors:** Chuqiao Li, Zhiwu Huang, Danda Pani Paudel, Yabin Wang, Mohamad Shahbazi, Xiaopeng Hong, Luc Van Gool60 61**Venue:** IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)62 63**Year:** 202364 65## Dataset Structure66 67This repository currently hosts the dataset archive:68 69- `CDDB.tar`70 71After extraction, the dataset is expected to contain benchmark splits and source-specific subsets used for continual deepfake detection experiments. According to the original paper and project repository, CDDB is built from a collection of real and manipulated images aggregated from multiple existing deepfake datasets and generation pipelines.72 73The benchmark includes deepfakes derived from multiple sources, including generative and manipulation pipelines such as:74 75- ProGAN76- StyleGAN77- BigGAN78- CycleGAN79- GauGAN80- CRN81- IMLE82- SAN83- FaceForensics++84- WhichFaceReal85- GLOW86- StarGAN87- WildDeepfake88 89The original benchmark is organized around different task sequences, including easy, hard, and long continual streams.90 91## Dataset Creation92 93CDDB was proposed to study continual deepfake detection in a more practical setting where deepfake generators evolve over time.94Instead of treating detection as a stationary benchmark, the dataset groups data into sequential tasks so that models can be evaluated on adaptation, retention, and generalization.95 96The benchmark is assembled from previously released open-source deepfake datasets and generation sources, rather than being collected from a single acquisition pipeline.97 98## Intended Uses99 100CDDB is intended for research use in:101 102- deepfake detection103- continual learning104- incremental learning105- robustness analysis under source shift106- benchmarking anti-forgetting strategies107 108It is particularly suitable for evaluating methods that must maintain performance on previously seen deepfake sources while adapting to newly introduced manipulations.109 110## Out-of-Scope Uses111 112This dataset is not intended to:113 114- certify production-ready deepfake detectors115- serve as a complete benchmark for all real-world manipulations116- support identity, biometric, or surveillance decisions117- be used in safety-critical or high-stakes automated decision systems without additional validation118 119## Considerations and Limitations120 121- CDDB is assembled from multiple existing datasets and generation methods, so its licensing and redistribution conditions may depend on the underlying sources.122- The benchmark reflects the manipulation methods and dataset availability at the time of the original publication.123- Performance on CDDB does not guarantee robustness to newer generative models or real-world post-processing pipelines.124- Models trained on this dataset may learn source-specific artifacts instead of general manipulation cues.125 126## Licensing Information127 128The license for this redistributed archive is currently marked as `unknown`.129Users should verify the licensing and redistribution terms of the original CDDB release and all upstream component datasets before commercial use or redistribution.130 131## Citation132 133If you use this dataset, please cite the original paper:134 135```bibtex136@InProceedings{Li_2023_WACV,137  author    = {Li, Chuqiao and Huang, Zhiwu and Paudel, Danda Pani and Wang, Yabin and Shahbazi, Mohamad and Hong, Xiaopeng and Van Gool, Luc},138  title     = {A Continual Deepfake Detection Benchmark: Dataset, Methods, and Essentials},139  booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},140  month     = {January},141  year      = {2023},142  pages     = {1339--1349}143}144```145 146## Acknowledgements147 148This dataset card is based on the original WACV 2023 paper and the official project repository. Credit for the benchmark, data construction, and experimental protocol belongs to the original authors.149