Geometry Cloak: Preventing TGS-based 3D Reconstruction from Copyrighted Images

Authors: Qi Song, Ziyuan Luo, Ka Chun Cheung, Simon See, Renjie Wan

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments have verified the effectiveness of our geometry cloak. Our project is available at https://qsong2001.github.io/geometry_cloak.
Researcher Affiliation Collaboration 1Department of Computer Science, Hong Kong Baptist University 2NVIDIA AI Technology Center, NVIDIA
Pseudocode Yes Algorithm 1: Optimizing Geometry Cloak with view-specific PGD
Open Source Code Yes Our project is available at https://qsong2001.github.io/geometry_cloak.
Open Datasets Yes To evaluate the performance of our method, we conduct experiments on the Google Scanned Objects (GSO) [7] and Omni Object3D (Omni3D) [45] datasets.
Dataset Splits No The paper mentions selecting images from datasets for evaluation but does not specify explicit training, validation, or test dataset splits (e.g., percentages or counts).
Hardware Specification Yes Our method uses the Py Torch framework on a single NVIDIA V100 GPU.
Software Dependencies No The paper mentions 'PyTorch framework' but does not specify a version number for it or any other software dependencies.
Experiment Setup Yes The input image I(0) = I is initialized, and iteratively updated for N = 100 steps with a step size of α = 0.001. The loss is defined as the Chamfer Distance [1] between the predicted point cloud P and the target point cloud ˆPtar. The updated image I(i) is clipped to the valid range [0, 1], and after N = 100 iterations, the final image ˆI = I(N) are obtained as the geometry cloaked output.