Kai-Yin-UoA/Monocular_Depth_Essentials
Monocular_Depth_Essentials Monocular_Depth_Essentials is a lightweight, curated core dataset tailored specifically for Monocular Depth Estimation tasks. Following the data preparation guidelines from the classic bts repository , this dataset extracts only the essential image-depth pairs from the massive raw KITTI and NYU Depth V2 datasets based strictly on the official Eigen Split train/test text lists. If you are benchmarking or reproducing Depth Anything (V1/V2), BTS, or… See the full description on the dataset page: https://huggingface.co/datasets/Kai-Yin-UoA/Monocular_Depth_Essentials.
MonocularDepthEssentials
Monocular_Depth_Essentials is a lightweight, curated core dataset tailored specifically for Monocular Depth Estimation tasks.
Following the data preparation guidelines from the classic bts repository , this dataset extracts only the essential image-depth pairs from the massive raw KITTI and NYU Depth V2 datasets based strictly on the official Eigen Split train/test text lists.
If you are benchmarking or reproducing Depth Anything (V1/V2), BTS, or other monocular depth estimation models, this dataset allows you to bypass the tedious raw data downloading, filtering, and cleaning phases, offering a true "plug-and-play" experience.
📌 Dataset Features & Structure
- Streamlined Data: Redundant sequences not evaluated in standard benchmarks are excluded, leaving only the exact samples specified by the official
txtsplits. - Standard Evaluation: Fully adheres to the academic standard Eigen Split, ensuring fair and direct experimental comparison.
- Clean Directory: Organised to seamlessly align with dataloaders in mainstream depth estimation codebases.