Awesome-AIGC-Image-Video-Detection

KO | EN

gitlite — search

#typescript #ai-agents #ai #dsh-plugin #deepseek-harness #open-source #claude-code #codex #cli #developer-tools #react #windows

Awesome-AIGC-Image-Video-Detection

★ 433 Open GitHub ↗

A curated collection of the latest research and resources on AI-Generated Image and Video Detection.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Awesome AIGC Image/Video Detection [![Awesome](https://awesome.re/badge.svg)](https://awesome.re)

A curated collection of the latest research and resources on AI-Generated Image and Video Detection. This repository encompasses datasets, benchmarks, research papers, and practical detection tools.

🚀🚀🚀Contributions are welcome! If you find any missing papers, datasets, or tools, feel free to open an issue or submit a pull request.

## Contents
- [Hot Events](#-hot-events)
- [Benchmarks & Datasets](#benchmarks--datasets)
- [Research Papers](#research-papers)
  - [MLLM-Based](#mllm-based)
  - [Classification-Based](#classification-based)
- [Competitions](#competitions)
- [Practical Detection Tools](#practical-detection-tools)
- [About Our Team](#-about-our-team)

---

## 🔥 Hot Events

- [一颗榴莲引发的1600公里追凶：谁在滥用“仅退款”？](https://h5.ifeng.com/c/vivo/v002Db8HLomgpDUJcI-_--LPT1YGzjdQXYG2RSCG3E68FhaZ0__?isNews=1&showComments=0)
- [因为GPT-image-2，整个互联网都变成了巨大的黑暗森林](https://mp.weixin.qq.com/s/zua1k53RovAOk15Juy6q3g)
- [ChatGPT 图像 2.0 现已上线! 开启图像生成的新纪元](https://openai.com/zh-Hans-CN/index/introducing-chatgpt-images-2-0/)
- [谷歌Gemini生图的暗水印，被印度小哥破解了，成功率90%，Github揽星3k+](https://github.com/aloshdenny/reverse-SynthID)
- [炸裂！技术力究极恐怖！米哈游蔡浩宇首款AI生成视频大模型LPM1.0曝光！自定义虚拟角色？支持多语言+无限真实+无限时长，实现自由对话、唱歌、表演等效果](https://www.bilibili.com/video/BV12SQwBEEZd/?vd_source=dc916dd94382a611d2fc7ba992618065)

- [⭐️ 刚刚，阿里官方认领神秘「欢乐马」(HappyHorse)，来自ATH郑波团队](https://weibo.com/8214551477/5286126636237040)

- [就AI换脸合成、影视素材魔改等，中广联演员委员会发声明](https://m.thepaper.cn/newsDetail_forward_32880368)

- [Is the new video of Netanyahu AI-generated?](https://x.com/ripplexrpie/status/2033212021647270063?s=46)

- [Netanyahu, disappearing ring!](https://x.com/jacksonhinklle/status/2033543544124633345?s=46)

- [为什么内塔尼亚胡没能成功证明自己没死？](https://www.bilibili.com/video/BV1ygwXzVEns?buvid=ZF4153F466FA276C4159AA3E46125C0C024C&from_spmid=tm.recommend.0.0&is_story_h5=false&mid=enL%2FASv4UwiQw63hS3bfYw%3D%3D&plat_id=114&share_from=ugc&share_medium=iphone&share_plat=ios&share_session_id=83DD9880-F093-4A78-8337-FAE59EB0F9A0&share_source=COPY&share_tag=s_i&timestamp=1773807790&unique_k=FyMOQcn&up_id=346563107&vd_source=516db97bb15e7a9ee84b3097ee2ff160)

- [A photo of Iran’s bombed schoolgirl graveyard went viral. Why did AI say it wasn’t real?](https://www.theguardian.com/global-development/2026/mar/17/atrocity-ai-slop-verify-facts-iran-minab-graves)

- [How AI Content Detection is Being Weaponized in the Iran War](https://www.techpolicy.press/how-ai-content-detection-is-being-weaponized-in-the-iran-war/)

- [改变视频行业的AI，快来了(但有点恐怖) - Bilibili 影视飓风](https://www.bilibili.com/video/BV1A3cczZEf6/?spm_id_from=333.1387.homepage.video_card.click&vd_source=516db97bb15e7a9ee84b3097ee2ff160)

- [勘误致歉：今日发布的库里采访资讯为AI杜撰生成](https://m.hupu.com/bbs-share/637055192.html?share=share&euid=7vQPQEPpuWG8GWXVjyxIRtUcrudfXWKbBiuRjnUu2JY=&cid=160373438)

- [聊聊关于Seedance 2.0的风波 - 数字生命卡兹克](https://mp.weixin.qq.com/s/FzHUB8uSQllJviMu6SLMew)

- [Trump unveils his vision for 'Gaza Riviera' with AI video featuring belly dancers and luxury yachts](https://www.youtube.com/shorts/ZIwqq3RDuc0)

- [电商上演「魔法对轰」：卖家用AI假图骗下单，买家拿AI烂水果骗退款](https://36kr.com/p/3409700220685704)

---

## Benchmarks & Datasets

> **Modality Legend:** `[I]` Image | `[V]` Video | `[M]` Multi-modal

> **Annotation Type Legend:** 
> `Au`: Authenticity | `Ex`: Explainability | `Lo`: Localization

⬆ [Back to Top](#contents)

---

## Research Papers

> **💡 Note:** Papers are sorted by year (descending) within each category. 
> **Modality Legend:** `[I]` Image | `[V]` Video | `[M]` Multi-modal

### MLLM-Based
*This category focuses on utilizing Multimodal Large Language Models (MLLMs) like GPT-4V, LLaVA, or Qwen-VL to detect AI-generated content. These methods often provide natural language explanations (explainability) alongside binary detection.*

| Title | Venue & Year | Modality | Highlights/Keywords | Code |
| --- | --- | --- | --- | --- |
| [Veritas++: Value-aware On-Policy Distillation for Perception-Enhanced AIGI Detection](https://arxiv.org/abs/2607.27113) | Arxiv 2026 | `[I]` | Value-aware On-Policy Distillation, Perception-Enhanced Reasoning | [GitHub](https://github.com/EricTan7/VeritasPP) |
| [Detecting AI-Generated Video: A Vision-Language Dual-View Survey](https://arxiv.org/abs/2607.10787) | ACL 2026 Findings | `[V]` | [Survey] Vision-Language Dual-View Taxonomy, Factual Fidelity Verification, Cross-modal Consistency | N/A |
| [TranX-Adapter: Bridging Artifacts and Semantics within MLLMs for Robust AI-generated Image Detection](https://arxiv.org/abs/2602.21716) | ICML 2026 | `[I]` | Artifact feature and Semantic feature fusion | [GitHub](https://github.com/DreamMr/TranX-Adapter) |
| [Venus-DeFakerOne: Unified Fake Image Detection & Localization](https://arxiv.org/pdf/2605.14091) | Arxiv 2026 | `[I]` | Unified Detection & Localization, Large-Scale Training | [GitHub](https://github.com/venus-guangjian/Venus-DeFakerOne) |
| [GenShield: Unified Detection and Artifact Correction for AI-Generated Images](https://arxiv.org/pdf/2605.16122) | ICML 2026 | `[I]` | Unified MLLM, Detect & Correct Artifacts | [GitHub](https://github.com/zhipeixu/GenShield) |
| [ReAlign: Generalizable Image Forgery Detection via Reasoning-Aligned Representation](https://arxiv.org/pdf/2605.16080) | Arxiv 2026 | `[I]` | Reasoning-Aligned Representation, Interpretable | N/A |
| [UniGenDet: A Unified Generative-Discriminative Framework for Co-Evolutionary Image Generation and Generated Image Detection](https://arxiv.org/abs/2604.21904v1) | CVPR 2026 | `[I]` | Unified Model, Co-Evolution (Generation & Detection)  | [GitHub](https://github.com/Zhangyr2022/UniGenDet) |
| [VideoVeritas: AI-Generated Video Detection via Perception Pretext Reinforcement Learning](https://arxiv.org/pdf/2602.08828) | ICML 2026 | `[V]` | Perception Pretext RL, Fact-based Reasoning, MintVid Dataset | [GitHub](https://github.com/EricTan7/VideoVeritas) |
| [Veritas: Generalizable deepfake detection via pattern-aware reasoning](https://arxiv.org/pdf/2508.21048) | ICLR 2026(Oral) | `[I]` | Pattern-aware Reasoning, HydraFake Dataset | [Github](https://github.com/EricTan7/Veritas) |
| [DF-LLaVA: Unlocking MLLMs for Synthetic Image Detection via Knowledge Injection and Conflict-Driven Self-Reflection](https://arxiv.org/pdf/2509.14957) | Arxiv 2026 | `[I]` | Knowledge Injection, Self-Reflection | N/A |
| [DocShield: Towards AI Document Safety via Evidence-Grounded Agentic Reasoning](https://arxiv.org/pdf/2604.02694) | Arxiv 2026 | `[M]` | Agentic Framework, Document Safety | N/A |
| [VidGuard-R1: AI-Generated Video Detection and Explanation via Reasoning MLLMs and RL](https://arxiv.org/abs/2510.02282) | ICLR 2026 | `[V]` | Multi-stage RL, Video Detection Dataset | [GitHub](https://github.com/kyoungjunpark/VidGuard-R1) |
| [FakeXplain: AI-Generated Image Detection via Human-Aligned Grounded Reasoning](https://arxiv.org/pdf/2506.07045) | ICLR 2026 | `[I]` | Grounded Reasoning, Human-annotated Dataset | N/A |
| [AlignGemini: Generalizable AI-Generated Image Detection Through Task-Model Alignment](https://arxiv.org/pdf/2512.06746) | Arxiv 2026 | `[I]` | Decoupling (Semantic & Pixel), AIGI-Now Dataset | N/A |
| [Zoom-In to Sort AI-Generated Images Out](https://arxiv.org/pdf/2510.04225) | Arxiv 2026 | `[I]` | Thinking with Images, MagniFake Dataset | N/A |
| [AgentFoX: LLM Agent-Guided Fusion with eXplainability for AI-Generated Image Detection](https://arxiv.org/pdf/2603.23115) | Arxiv 2026 | `[I]` | Agentic framework | [Github](https://github.com/suncore946/AgentFoX) |
| [EvoGuard: An Extensible Agentic RL-based Framework for Practical and Evolving AI-Generated Image Detection](https://arxiv.org/pdf/2603.17343) | Arxiv 2026 | `[I]` | Agentic Framework, Method Ensembling | N/A |
| [VIGIL: Part-Grounded Structured Reasoning for Generalizable Deepfake Detection](https://arxiv.org/pdf/2603.21526) | Arxiv 2026 | `[I]` | Part-centric Forensic, OmniFake Dataset | [Project](https://vigil.best/) |
| [GenVideoLens: Where LVLMs Fall Short in AI-Generated Video Detection?](https://arxiv.org/pdf/2603.18625) | Arxiv 2026 | `[V]` | GenVideoLens benchmark | N/A |
| [Semantic Visual Anomaly Detection and Reasoning in AI-Generated Images](https://arxiv.org/pdf/2510.10231) | ICLR 2026 | `[I]` | Semantic Anomaly Reasoning, AnomReason Dataset | N/A |
| [FAKE-HR1: RETHINKING REASONING OF VISION LANGUAGE MODEL FOR SYNTHETIC IMAGE DETECTION](https://arxiv.org/pdf/2602.10042v1) | Arxiv 2026 | `[I]` | Hybrid-Reasoning, Dual-mode Dataset | N/A |
| [MIRAGE: Towards AI-Generated Image Detection in the Wild](https://arxiv.org/pdf/2508.13223) | Arxiv 2025 | `[I]` | Human Curation Dataset, Heuristic-to-Analytic Reasoning | N/A |
| [BusterX++: Towards Unified Cross-Modal AI-Generated Content Detection and Explanation with MLLM](https://www.alphaxiv.org/pdf/2507.14632) | Arxiv 2025 | `[M]` | RL Post-training, Cross-Modal, Thinking Reward Mechanism | [Github](https://github.com/l8cv/BusterX) |
| [BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation](https://www.alphaxiv.org/pdf/2505.12620) | Arxiv 2025 | `[V]` | GenBuster-200K Dataset, Cold Start + RL Training | [Github](https://github.com/l8cv/BusterX) |
| [REVEAL: Reasoning-enhanced Forensic Evidence Analysis for Explainable AI-generated Image Detection](https://arxiv.org/pdf/2511.23158) | Arxiv 2025 | `[I]` | Chain-of-Evidence, Expert-grounded RL | N/A |
| [Spot the Fake: Large Multimodal Model-Based Synthetic Image Detection with Artifact Explanation](https://arxiv.org/pdf/2503.14905) | NeurIPS 2025 | `[I]` | FakeClue Dataset, Fine-grained Artifact Clues, Artifact Explanation | [GitHub](https://github.com/opendatalab/FakeVLM) |
| [AIGI-Holmes: Towards Explainable and Generalizable AI-Generated Image Detection via Multimodal Large Language Models](https://arxiv.org/pdf/2507.02664) | ICCV 2025 | `[I]` | Holmes-Set, Multi-Expert Jury, 3-Stage Training Pipeline | [Github](https://github.com/wyczzy/AIGI-Holmes) |
| [LEGION: Learning to Ground and Explain for Synthetic Image Detection](https://arxiv.org/pdf/2503.15264) | ICCV 2025 | `[I]` | SynthScars Dataset, Defender & Controller, Image Refinement | [GitHub](https://github.com/opendatalab/LEGION) |
| [Seeing Before Reasoning: A Unified Framework for Generalizable and Explainable Fake Image Detection](https://arxiv.org/pdf/2509.25502) | Arxiv 2025 | `[I]` | Perception & Reasoning, ExplainFake-Bench | N/A |
| [SIDA: Social Media Image Deepfake Detection, Localization, and Explanation](https://arxiv.org/pdf/2412.04292) | CVPR 2025 | `[I]` | SID-Set, Mask Prediction, Social Media Context | [Github](https://github.com/hzlsaber/SIDA) |
| [FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models](https://arxiv.org/pdf/2410.02761) | ICLR 2025 | `[I]` | Explainable IFDL, Domain Tag-guided, Multi-modal Localization | [GitHub](https://github.com/zhipeixu/FakeShield) |
| [VidGuard-R1: AI-Generated Video Detection and Explanation via Reasoning MLLMs and RL](https://arxiv.org/pdf/2510.02282) | ICLR 2026 | `[V]` | GRPO, Time Artifacts, Quality Evolutionary Videos | N/A |
| [FakeScope: Large Multimodal Expert Model for Transparent AI-Generated Image Forensics](https://arxiv.org/pdf/2503.24267) | Arxiv 2025 | `[I]` | FakeChain Dataset, FakeInstruct, Trace Evidence | N/A |
| [AntifakePrompt: Prompt-Tuned Vision-Language Models are Fake Image Detectors](https://arxiv.org/pdf/2310.17419) | Arxiv 2024 | `[I]` | VQA, InstructBLIP, Soft Prompt-tuning, Zero-shot | [GitHub](https://github.com/nctu-eva-lab/AntifakePrompt) |

### Classification-Based
*This category includes supervised learning approaches that train neural networks (CNNs, ViTs, etc.) specifically to classify authentic vs. AI-generated content. They usually focus on robustness, generalization, and feature extraction.*

| Title | Venue & Year | Modality | Highlights/Keywords | Code |
| --- | --- | --- | --- | --- |
| [Rethinking the Readout: Unlocking Video Backbones for AI-Generated Video Detection](https://arxiv.org/abs/2607.15321) | Arxiv 2026 | `[V]` | V-PVP Readout, Patch Velocity Profiling, Frozen Video Backbones | [Code](https://anonymous.4open.science/r/PVP-81B3/) |
| [GlobalForge: Towards Robust AI-Generated Image Detection](https://arxiv.org/abs/2607.14684) | Arxiv 2026 | `[I]` | Global Structural Reasoning, Local Information Bottleneck, RealDeg-Bench | [Code](https://anonymous.4open.science/r/GlobalForge-BE0F/) |
| [SPLIT: Training-Free AI-Generated and Partially Edited Video Detection via Spatial Patch-Level Incoherence and Temporal Roughness](https://arxiv.org/abs/2607.02886) | ECCV 2026 | `[V]` | Training-free, Patch-level Incoherence, Temporal Roughness, Ultra-low FPR | [GitHub](https://github.com/mldljyh/SPLIT) |
| [Dataset Biases and Shortcut Learning in Motion-Based AI-Generated Video Detection](https://arxiv.org/abs/2607.00948) | Arxiv 2026 | `[V]` | Motion Bias Analysis, Preprocessing/Sampling Bias, Frequency-based Comparison | N/A |
| [G2VD: Generalizable AI-Generated Video Detection via Counterfactual Intervention and Causal Disentanglement](https://arxiv.org/abs/2607.04607) | Arxiv 2026 | `[V]` | Counterfactual Intervention, Causal Disentanglement, Cross-domain Generalization | [GitHub](https://github.com/dumeng98/G2VD) |
| [Fleet: Few Shots Lead Effective AI-generated Image Detection](https://arxiv.org/abs/2606.31082) | ICML 2026 | `[I]` | Few-shot Adaptation, Routing Correction, Treasure Benchmark | [GitHub](https://github.com/ICTMCG/Fleet) |
| [SSAFE: Simple and Strong AI-Generated Image Detection via Frozen Vision Encoders](https://arxiv.org/abs/2606.08634) | Arxiv 2026 | `[I]` | Frozen Vision Encoders, Linear Classifier, RealWorldBench | N/A |
| [When Seeing Is Not Believing -- A Benchmark for Search-Grounded Video Misinformation Detection](https://arxiv.org/pdf/2606.04098) | Arxiv 2026 | `[V]` | Search-Grounded Verification, EVID-Bench, Evidence-Dependent Manipulation | N/A |
| [ReConFuse: Reconstruction-Error Guided Semantic Fusion for AI-Generated Video Detection](https://arxiv.org/pdf/2606.04706) | Arxiv 2026 | `[V]` | Reconstruction Error, Semantic Fusion, Spatial-Temporal Artifacts | N/A |
| [HydraPrompt: An Adaptive and Asymmetric Framework of Vision-Language Models for Synthetic Image Detection](https://arxiv.org/pdf/2605.26421) | ACM MM 2026 | `[I]` | Asymmetric Prompting, Dynamic Decision Boundary | N/A |
| [VINA: Video as Natural Augmentation: Towards Unified AI-Generated Image and Video Detection](https://arxiv.org/pdf/2605.21977) | Arxiv 2026 | `[M]` | Unified Image/Video Detection, Cross-Modal Contrastive Learning | N/A |
| [Training-free Detection of Generated Videos via Spatial-Temporal Likelihoods](https://arxiv.org/pdf/2603.15026) | CVPR 2026 | `[V]` | Training-free, Zero-shot, Spatial-Temporal Likelihoods, ComGenVid Dataset | [GitHub](https://github.com/OmerBenHayun/STALL) |
| [PGC: Peak-Guided Calibration for Generalizable AI-Generated Image Detection](https://arxiv.org/pdf/2605.21207) | ICML 2026 | `[I]` | Peak-Guided Calibration, CommGen15 Dataset | [GitHub](https://github.com/xiaoyu6868/PGC) |
| [Reduce the Artifacts Bias for More Generalizable AI-Generated Image Detection](https://arxiv.org/pdf/2605.14486) | Arxiv 2026 | `[I]` | Bias-free Training | [GitHub](https://github.com/liyih/SEF_AIGC_detection) |
| [Detecting AI-Generated Videos with Spiking Neural Networks](https://arxiv.org/pdf/2605.05895) | Arxiv 2026 | `[V]` | Spiking Neural Networks, Temporal Artifact | N/A |
| [Robust ASIC-Based Image Authentication Using Reed-Solomon LSB Watermarking](https://github.com/Agnuxo1/Secure_image_generation_with_ASIC_signature) | Preprint 2026 | `[I]` | ASIC PoW, Hardware-bound Provenance, Reed-Solomon Watermarking | [GitHub](https://github.com/Agnuxo1/Secure_image_generation_with_ASIC_signature) |
| [Automated In-the-Wild Data Collection for Continual AI Generated Image Detection](https://arxiv.org/pdf/2605.02567) | Arxiv 2026 | `[I]` | Continual Learning, Continual Data Collection | [GitHub](https://mever-team.github.io/WildFC/) |
| [IncreFA: Breaking the Static Wall of Generative Model Attribution](https://arxiv.org/pdf/2604.17736) | Arxiv 2026 | `[I]` | Incremental Learning, Generative Model Attribution | [GitHub](https://github.com/Ant0ny44/IncreFA) |
| [CMTA: Leveraging Cross-Modal Temporal Artifacts for Generalizable AI-Generated Video Detection](https://arxiv.org/pdf/2605.00630) | Arxiv 2026 | `[V]` | Cross-Modal Temporal Artifacts, Video Detection | N/A |
| [SAIDO: Generalizable Detection of AI-Generated Images via Scene-Aware and Importance-Guided Dynamic Optimization in Continual Learning](https://arxiv.org/pdf/2512.00539) | CVPR 2026 | `[I]` | Scene-aware optimization, Continual learning | [GitHub](https://github.com/edu-yinzhaoxia/SAIDO-CVPR2026) |
| [Zooming In on Fakes: A Novel Dataset for Localized AI-Generated Image Detection with Forgery Amplification Approach](https://arxiv.org/abs/2504.11922) | AAAI 2026 | `[I]` | Localized AIGC Detection, Forgery Amplification, Scene-aware Local Forgery | [GitHub](https://github.com/clpbc/BR-Gen) |
| [Preserving Forgery Artifacts: AI-Generated Video Detection at Native Scale](https://openreview.net/pdf?id=XD43lfRCg6) | ICLR 2026 | `[V]` | Native scale video processing, Massive realistic video dataset, Preserves subtle generation artifacts | N/A |
| [Generalizable and Adaptive Continual Learning Framework for AI-generated Image Detection](https://arxiv.org/pdf/2601.05580) | TMM 2026 | `[I]` | Continual Learning, Kronecker-Factored Approximate Curvature | N/A |
| [Simplicity Prevails: The Emergence of Generalizable AIGI Detection in Visual Foundation Models](https://arxiv.org/pdf/2602.01738) | Arxiv 2026 | `[I]` | Linear Probe, Vision Foundation Models, Emergent Forensic Capability | N/A |
| [MIRROR: Manifold Ideal Reference ReconstructOR for Generalizable AI-Generated Image Detection](https://arxiv.org/pdf/2602.02222) | Arxiv 2026 | `[I]` | Manifold Reconstruction, Memory Bank, Human-AIGI Benchmark | [GitHub](https://github.com/349793927/MIRROR) |
| [No Pixel Left Behind: A Detail-Preserving Architecture for Robust High-Resolution AI-Generated Image Detection](https://arxiv.org/pdf/2508.17346) | ICLR 2026 | `[I]` | Detail-preserving dual-path architecture, Multi-task learning, HiRes-50K benchmark  | N/A |
| [All Patches Matter, More Patches Better: Enhance AI-Generated Image Detection via Panoptic Patch Learning](https://arxiv.org/pdf/2504.01396) | ICLR 2026 | `[I]` | Random Patch Replacement, Patch-wise Contrastive Learning  | N/A |
| [OmniAID: Decoupling Semantic and Artifacts for Universal AI-Generated Image Detection in the Wild](https://arxiv.org/pdf/2511.08423) | Arxiv 2025 | `[I]` | Mixture-of-Experts, Semantic-Artifact Decoupling, Mirage Dataset | [GitHub](https://github.com/yunncheng/OmniAID) |
| [DINO-Detect: A Simple yet Effective Framework for Blur-Robust AI-Generated Image Detection](https://arxiv.org/pdf/2511.12511) | Arxiv 2025 | `[I]` | Blur Robustness, Knowledge Distillation, DINOv3 | [Github](https://github.com/JiaLiangShen/Dino-Detect-for-blur-robust-AIGC-Detection) |
| [D3QE: Learning Discrete Distribution Discrepancy-aware Quantization Error for Autoregressive-Generated Image Detection](https://openaccess.thecvf.com/content/ICCV2025/papers/Zhang_D3QE_Learning_Discrete_Distribution_Discrepancy-aware_Quantization_Error_for_Autoregressive-Generated_Image_ICCV_2025_paper.pdf) | ICCV 2025 | `[I]` | Discrete Distribution Discrepancy-aware Transformer, Vector Quantized Variational AutoEncoder | [Github](https://github.com/Zhangyr2022/D3QE) |
| [Seeing What Matters: Generalizable AI-generated Video Detection with Forensic-Oriented Augmentation](https://arxiv.org/pdf/2506.16802) | NeurIPS 2025 | `[V]` | Wavelet-band Augmentation, Forensic Frequency Artifacts, Single-generator Generalization | [GitHub](https://github.com/grip-unina/WaveRep-SyntheticVideoDetection) |
| [AI-Generated Video Detection via Perceptual Straightening](https://arxiv.org/pdf/2507.00583) | NeurIPS 2025 | `[V]` | Perceptual Straightening, DINOv2, Temporal Curvature | [GitHub](https://github.com/ChristianInterno/ReStraV) |
| [Physics-Driven Spatiotemporal Modeling for AI-Generated Video Detection](https://arxiv.org/pdf/2510.08073) | NeurIPS 2025 | `[V]` | Normalized Spatiotemporal Gradient (NSG), Maximum Mean Discrepancy (MMD) | [Github](https://github.com/ZSHsh98/NSG-VD) |
| [Dual Data Alignment Makes AI-Generated Image Detector Easier Generalizable](https://arxiv.org/pdf/2505.14359) | NeurIPS 202 (Spotlight) | `[I]` | Dual-domain Alignment, Frequency-level Bias, VAE Reconstruction | [GitHub](https://github.com/roy-ch/Dual-Data-Alignment) |
| [Orthogonal Subspace Decomposition for Generalizable AI-Generated Image Detection](https://arxiv.org/pdf/2411.15633) | ICML 2025 (Oral) | `[I]` | SVD Orthogonal Subspace, Asymmetry Phenomenon, Parameter-efficient Fine-tuning | [GitHub](https://github.com/zhiyuan-yan/Effort) |
| [Any-Resolution AI-Generated Image Detection by Spectral Learning](https://arxiv.org/pdf/2411.19417) | CVPR 2025 | `[I]` | Spectral Context Attention, Frequency Reconstruction, OOD Detection | [Github](https://github.com/kartyg23/spai) |
| [A Bias-Free Training Paradigm for More General AI-generated Image Detection](https://arxiv.org/pdf/2412.17671) | CVPR 2025 | `[I]` | Bias-Free, Semantic Alignment, Stable Diffusion Self-conditioning | [Github](https://github.com/grip-unina/B-Free) |
| [Forensics Adapter: Adapting CLIP for Generalizable Face Forgery Detection](https://arxiv.org/pdf/2411.19715) | CVPR 2025 | `[I]` | CLIP, Blending Boundaries, Forgery-aware Prompt Learning | [Github](https://github.com/OUC-VAS/ForensicsAdapter) |
| [Towards a Universal Synthetic Video Detector: From Face or Background Manipulations to Fully AI-Generated Content](https://arxiv.org/pdf/2412.12278) | CVPR 2025 | `[V]` | SigLIP-So400M, Attention-Diversity Loss, Full-frame Manipulations | N/A |
| [Exploring Unbiased Deepfake Detection via Token-Level Shuffling and Mixing](https://arxiv.org/pdf/2501.04376) | AAAI 2025 | `[I]` | Token-Level Shuffling, Contrastive Loss, Bias Mitigation | N/A |
| [DIP: Diffusion Learning of Inconsistency Pattern for General DeepFake Detection](https://arxiv.org/pdf/2410.23663) | TMM 2025 | `[V]` | Direction-aware Attention, SpatioTemporal Invariant Loss | N/A |
| [Frequency-Aware Deepfake Detection: Improving Generalizability through Frequency Space Domain Learning](https://arxiv.org/pdf/2403.07240) | AAAI 2024 | `[I]` | Frequency Domain, FFT, Frequency Conv Layer (FCL), Lightweight | [GitHub](https://github.com/chuangchuangtan/FreqNet-DeepfakeDetection) |
| [DeMamba: AI-Generated Video Detection on Million-Scale GenVideo Benchmark](https://arxiv.org/pdf/2405.19707) | Arxiv 2024 | `[V]` | Mamba, State Space Model, Long-range Spatiotemporal Inconsistency | [GitHub](https://github.com/chenhaoxing/DeMamba) |
| [Rethinking the Up-Sampling Operations in CNN-based Generative Network](https://arxiv.org/pdf/2312.10461) | CVPR 2024 | `[I]` | Neighboring Pixel Relationships, Generalized Structural Artifacts | [Github](https://github.com/chuangchuangtan/NPR-DeepfakeDetection) |
| [Distinguish Any Fake Videos: Unleashing the Power of Large-scale Data and Motion Features](https://arxiv.org/pdf/2405.15343) | Arxiv 2024 | `[V]` | GenVidDet, Optical Flow, Dual-Branch 3D Transformer | N/A |
| [FakeFormer: Efficient Vulnerability-Driven Transformers for Generalisable Deepfake Detection](https://arxiv.org/pdf/2410.21964) | Arxiv 2024 | `[I]` | Vulnerability-driven, Local Attention (L2-Att), Vision Transformer | [GitHub](https://github.com/10Ring/FakeFormer) |

⬆ [Back to Top](#contents)

---

## Competitions

| Competition | Link | Year | Info |
| :---------- | :--- | :--- | :--- |
| Robust AIGC Detection | [NTIRE 2026 Robust AI-Generated Image Detection in the Wild](https://www.codabench.org/competitions/12761/#/pages-tab) | 2026 | No restrictions on training data. Evaluate ROC AUC metrics on robust samples. |
| Robust Deepfake Detection | [NTIRE 2026 Robust Deepfake Detection Challenge](https://www.codabench.org/competitions/12795/) | 2026 | No restrictions on training data. |
| The 6th Face Anti-spoofing Challenge | [The 6th Face Anti-Spoofing: Unified Physical-Digital Attacks Detection@ICCV2025](https://codalab.lisn.upsaclay.fr/competitions/22915) | 2025 | No external data or pre-trained models allowed. Limited to a single DL model with under 100G FLOPs. |
| Detect AI vs. Human-Generated Images | [2025 Women in AI (WAI) Kaggle Challenge](https://www.kaggle.com/competitions/detect-ai-vs-human-generated-images) | 2025 | Paired dataset of authentic and AI-generated images |
| The 5th Face Anti-spoofing Challenge | [5th Chalearn Face Anti-spoofing Workshop and Challenge@CVPR2024](https://sites.google.com/view/face-anti-spoofing-challenge/welcome/challengecvpr2024) | 2024 | UniAttackData+ for unified physical and digital attack detection. |

⬆ [Back to Top](#contents)

---

## Practical Detection Tools

- **SiliconSignature** - [GitHub](https://github.com/Agnuxo1/Secure_image_generation_with_ASIC_signature) - Hardware-bound image authentication using ASIC PoW nonces for unforgeable provenance certification
- **EyeSift** - [Website](https://www.eyesift.com/) - Free online AI text/image/video/audio detector with detailed per-model benchmarks
- **Hive Moderation** - [Website](https://thehive.ai/demos/ai-generated-content-detection)
- **Tencent Zhuque AI Detection Assistant** - [Website](https://matrix.tencent.com/ai-detect/ai_gen_txt)
- **AI or Not** - [Website](https://www.aiornot.com/)
- **Illuminarty** - [Website](https://app.illuminarty.ai/)
- **Winston AI** - [Website](https://gowinston.ai/ai-image-detector/)
- **Is it AI?** - [Website](https://isitai.com/ai-image-detector)
- **TruthScan** - [Website](https://truthscan.com/zh)
- **中科睿鉴 (Zhongke Ruijian)** - 微信小程序搜索 **睿鉴AI**

⬆ [Back to Top](#contents)

---

## 🏢 About Our Team
We are the **Content Security Intelligence** Team under **Ant Group - Machine Intelligence**. We are responsible for developing comprehensive content security and risk-mitigation capabilities for the Ant Group ecosystem, bridging the gap between rapidly evolving technologies and the urgent need for digital trust.

### Why We Do It
In an era where synthetic media is increasingly sophisticated and pervasive, our research serves as a critical line of defense. By advancing AIGC detection technologies, we aim to:
*   **Safeguard Digital Integrity:** We provide essential defense mechanisms to protect the authenticity of visual content and combat the spread of misinformation in the digital space.
*   **Empower Trust:** Our solutions ensure the public can distinguish between genuine and synthetic media, fostering a more transparent and trustworthy digital ecosystem.
*   **Industrial Application & Impact:** We provide robust, scalable aigc detection solutions for Ant Group’s diverse content platforms, including  **[Lingguang](https://www.lingguang.com/chat)**, **[Jingtan](https://jingtanbusiness.antgroup.com/index)**, and many others.

### 🤝 Collaborators
We are honored to collaborate with esteemed researchers and scholars in the field of AI and Computer Vision. We deeply value these academic partnerships that drive our innovation:

*   **Prof. Jun Wan (万军)** | [CASIA](http://www.ia.cas.cn/) & [UCAS](https://www.ucas.ac.cn/)
    - *Research Interests:* Biometrics, Face Anti-spoofing, Gesture Recognition, and Computer Vision.
    - [[Homepage]](https://people.ucas.edu.cn/~jwan)
*   **Prof. Jianfu Zhang (张健夫)** | [Shanghai Jiao Tong University](https://www.sjtu.edu.cn/)
    - *Research Interests:* Computer Vision, Pattern Recognition, and Image/Video Analysis & Synthesis.
    - [[Homepage]](https://www.cs.sjtu.edu.cn/jiaoshiml/zhangjianfu.html)
*   **Prof. Zhuosheng Zhang (张倬胜)** | [Shanghai Jiao Tong University](https://www.sjtu.edu.cn/)
    - *Research Interests:* Natural Language Processing, Large Language Models, and Multi-modal Learning.
    - [[Homepage]](https://www.cs.sjtu.edu.cn/jiaoshiml/zhangzhuosheng.html)

### 📝 Academic Publications
*   **Veritas++: Value-aware On-Policy Distillation for Perception-Enhanced AIGI Detection** | *arXiv, 2026*
    -   *Highlights:* Strengthened fine-grained visual perception for explainable AIGI detection through perception-oriented learning and value-aware on-policy distillation.
    -   [[Paper]](https://arxiv.org/abs/2607.27113) [[Code]](https://github.com/EricTan7/VeritasPP)
*   **VideoVeritas: AI-Generated Video Detection via Perception Pretext Reinforcement Learning** | *ICML'26, 2026*
    -   *Highlights:* Detected AI-generated videos using perception pretext reinforcement learning to capture temporal inconsistencies.
    -   [[Paper]](https://arxiv.org/pdf/2602.08828) [[Code]](https://github.com/EricTan7/VideoVeritas)
*   **Locate-Then-Examine: Grounded Region Reasoning Improves Detection of AI-Generated Images** | *CVPR'26, 2026*
    -   *Highlights:* Improved detection accuracy through a two-stage approach of localizing suspicious regions followed by detailed examination.
    -   [[Code]](https://github.com/Gennadiyev/LocateThenExamine)
*   **GAMMA: Generalizable Alignment via Multi-task and Manipulation-Augmented Training for AI-Generated Image Detection** | *ICASSP'26, 2026*
    -   *Highlights:* Enhanced generalization through multi-task learning and manipulation-augmented training strategies.
    -   [[Paper]](https://arxiv.org/pdf/2509.10250)
*   **FakeXplain: AI-Generated Image Detection via Human-Aligned Grounded Reasoning** | *ICLR'26, 2026*
    -   *Highlights:* Detected AI-generated images through human-aligned grounded reasoning, providing interpretable visual evidence.
    -   [[Paper]](https://openreview.net/pdf?id=UcpTOa8OnG) [[Code]](https://github.com/Gennadiyev/FakeXplain)
*   **Veritas: Generalizable deepfake detection via pattern-aware reasoning** | *ICLR'26 Oral, 2026*
    -   *Highlights:* Achieved generalizable deepfake detection through pattern-aware reasoning, improving robustness across diverse manipulation types.
    -   [[Paper]](https://arxiv.org/pdf/2508.21048) [[Code]](https://github.com/EricTan7/Veritas)
*   **Generalizable and Adaptive Continual Learning Framework for AI-generated Image Detection** | *IEEE TMM, 2025*
    -   *Highlights:* Proposed a continual learning framework that adapts to new generative models while mitigating catastrophic forgetting.
    -   [[Paper]](https://arxiv.org/pdf/2601.05580)
*   **Towards explainable fake image detection with multi-modal large language models** | *ACM MM'25, 2025*
    -   *Highlights:* Leveraged multi-modal large language models to provide human-interpretable explanations for fake image detection.
    -   [[Paper]](https://dl.acm.org/doi/pdf/10.1145/3746027.3755421)
*   **WildFake: A Large-scale Challenging Dataset for AI-Generated Images Detection** | *AAAI'25 Oral, 2024*
    -   *Highlights:* Introduced the largest and most comprehensive AIGC image dataset at the time, providing a challenging benchmark for detection models.
    -   [[Paper]](https://arxiv.org/pdf/2402.11843)

### 🏆 Competition Achievements
*   **1st Place Winner** | *NTIRE 2026 Robust AI-Generated Image Detection in the Wild Challenge*
    -   Secured the top rank in ROC AUC for delivering superior performance in large-scale, real-world AI-generated image detection.
    -   [[Challenge Website]](https://www.codabench.org/competitions/12761/#/pages-tab)
*   **1st Place Winner** | *ICCV 2025 VQualA Challenge - Image Super-Resolution Generated Content Quality Assessment, 2025*
    -   Achieved top performance in the VQualA 2025 challenge focused on assessing the quality of super-resolution generated content.
    -   [[Paper 1]](https://openaccess.thecvf.com/content/ICCV2025W/VQualA/papers/Li_VQualA_2025_Challenge_on_Image_Super-Resolution_Generated_Content_Quality_Assessment_ICCVW_2025_paper.pdf) [[Paper
2]](https://openaccess.thecvf.com/content/ICCV2025W/VQualA/papers/Li_Hybrid_Vision_Transformer_and_Convolutional_Neural_Network_for_Super-Resolution_Image_ICCVW_2025_paper.pdf)
*   **1st Place Winner** | *CVPR 2024 Face Anti-Spoofing Challenge, 2024*
    -   Secured first place in the prestigious Face Anti-Spoofing Challenge at CVPR 2024, demonstrating state-of-the-art detection capabilities.
    -   [[Challenge Website]](https://sites.google.com/view/face-anti-spoofing-challenge/welcome/challengecvpr2024)

### 🛠️  Open-Source Resources
*   **WildFake** - A large and comprehensive AIGC image detection dataset.
    -   [[ModelScope]](https://modelscope.cn/datasets/hy2628982280/WildFake)
*   **GenVideo** - A large and comprehensive AIGC video detection dataset.
    -   [[ModelScope]](https://modelscope.cn/datasets/cccnju/Gen-Video)
*   **HydraFake** - A large-scale challenging dataset for AI-generated image detection.
    -   [[ModelScope]](https://www.modelscope.cn/datasets/EricTanh/HydraFake)
*   **MintVid** - A comprehensive video dataset for AIGC detection research.
    -   [[ModelScope]](https://www.modelscope.cn/datasets/EricTanh/MintVid)

### ✉️ Contact Us
For questions or collaborations, please contact:

- Zijian Yu: yuzijian.yzj@antgroup.com
- Hao Tan: tanhao2023@ia.ac.cn
- Jun Lan: yelan.lj@antgroup.com

⬆ [Back to Top](#contents)

---

### Star History

---

Awesome-AIGC-Image-Video-Detection

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README