The repository provides the DL3DV-10K dataset, a large-scale collection of real-world videos annotated with scene information and camera parameters. It is designed to support research in novel view synthesis, 3D reconstruction, and 3D Gaussian splatting. The dataset is widely used by major AI projects including NVIDIA Cosmos (World Foundation Model training) and Stability AI (camera control video generation), and is available on Hugging Face with multiple processed variants.