CopyRNeRF: Protecting the CopyRight of Neural Radiance Fields
About
Neural Radiance Fields (NeRF) have the potential to be a major representation of media. Since training a NeRF has never been an easy task, the protection of its model copyright should be a priority. In this paper, by analyzing the pros and cons of possible copyright protection solutions, we propose to protect the copyright of NeRF models by replacing the original color representation in NeRF with a watermarked color representation. Then, a distortion-resistant rendering scheme is designed to guarantee robust message extraction in 2D renderings of NeRF. Our proposed method can directly protect the copyright of NeRF models while maintaining high rendering quality and bit accuracy when compared among optional solutions.
Ziyuan Luo, Qing Guo, Ka Chun Cheung, Simon See, Renjie Wan• 2023
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Digital Watermarking | Blender and LLFF (test) | Bit Accuracy (No Attack)91.16 | 15 | |
| 3D Scene Watermarking | Blender and LLFF 16 bits | Bit Accuracy91.16 | 14 | |
| 3D Scene Watermarking | Blender and LLFF 32 bits | Bit Accuracy78.08 | 14 | |
| 3D Scene Watermarking | Blender and LLFF 48 bits | Bit Acc60.06 | 14 | |
| 3D Watermarking Robustness against Diffusion Attacks | Blender and LLFF (test) | Bit Accuracy (Deterministic)0.512 | 6 | |
| Image Quality Assessment | Blender and LLFF views (test) | SSIM0.747 | 6 | |
| 3D Watermarking | LLFF and Blender (train) | Training Time (min)85 | 6 | |
| 3D Scene Watermarking and Reconstruction | Blender | PSNR30.29 | 5 | |
| 3D Scene Watermarking and Reconstruction | LLFF | PSNR24.03 | 5 | |
| 3D Scene Watermarking and Reconstruction | MipNeRF360 | PSNR22.47 | 5 |
Showing 10 of 10 rows