Summary of Decof: Generated Video Detection Via Frame Consistency: the First Benchmark Dataset, by Long Ma et al.
DeCoF: Generated Video Detection via Frame Consistency: The First Benchmark Dataset
by Long Ma, Jiajia Zhang, Hongping Deng, Ningyu Zhang, Qinglang Guo, Haiyang Yu, Yong Liao, Pengyuan Zhou
First submitted to arxiv on: 3 Feb 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed paper tackles the pressing issue of generated video detection, a crucial problem in today’s world where advanced video generation methods are becoming increasingly sophisticated. To address this challenge, the authors present an open-source dataset and a novel detection method called DeCoF (detection model based on frame consistency). The dataset consists of 964 prompts covering various scenarios, behaviors, and actions, as well as different generation models, including commercial ones like OpenAI’s Sora and Google’s Veo. The paper also highlights the limitations of spatial artifact-based detectors, which lack generalizability. DeCoF, on the other hand, focuses on temporal artifacts by eliminating spatial artifacts during feature learning. Experimental results demonstrate the effectiveness of DeCoF in detecting generated videos from unseen models and its ability to generalize across multiple proprietary models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about finding ways to tell real videos apart from fake ones made using advanced computer programs. Right now, these fake videos are hard to detect, which makes them a big security problem. The authors created a special dataset with lots of examples of different types of fake and real videos, as well as different computer models that make the fake videos. They also developed a new way to detect fake videos called DeCoF, which looks at how the video changes from one frame to another instead of just looking at the individual frames. This helps DeCoF to be more accurate and able to spot fake videos even if it’s never seen them before. |
