SeMantic InDustry
SeMantic InDustry (S.MID) is a dataset designed to advance the field of LiDAR semantic segmentation, specifically for robotic applications and large-scale industrial scene. The dataset is based on a hybrid-solid LiDAR (Livox Mid-360). To create S.MID, researchers used an industrial robot to collect a total of 38,904 frames of LiDAR data at a rate of 10 Hz across various substations. The LiDAR point clouds are annotated into 25 categories under professional guidance (14 categories for single frame segmentation task) .
Variants: S.MID
This dataset is used in 1 benchmark:
| Task | Model | Paper | Date |
|---|---|---|---|
| LIDAR Semantic Segmentation | SFPNet | SFPNet: Sparse Focal Point Network … | 2024-07-16 |
| LIDAR Semantic Segmentation | SphereFormer | Spherical Transformer for LiDAR-based 3D … | 2023-03-22 |
| LIDAR Semantic Segmentation | Cylinder3D | Cylindrical and Asymmetrical 3D Convolution … | 2020-11-19 |
| LIDAR Semantic Segmentation | Submanifold Sparse CNN | 3D Semantic Segmentation with Submanifold … | 2017-11-28 |
Recent papers with results on this dataset: