| pretty_name: Rubber Tree 3D LiDAR Point Cloud Dataset | |
| tags: | |
| - lidar | |
| - 3d-point-cloud | |
| - robotics | |
| - agriculture | |
| # Rubber Tree 3D LiDAR Point Cloud Dataset | |
| This repository contains a Gazebo-simulated 3D LiDAR point-cloud dataset for rubber-tree detection in agricultural robotics. | |
| The complete dataset contains 7,517 frames and 126,926 annotated tree instances from two simulated plantation layouts. | |
| ## Files | |
| - `S5K.zip`: curved-row plantation layout. | |
| - `straight_line.zip`: straight/grid-row plantation layout. | |
| The two original data archives are preserved unchanged. | |
| ## Evaluation protocol | |
| The complete S5K subset contains 5,545 frames and is used for training and validation/model selection: | |
| - Training: 4,436 S5K frames. | |
| - Validation/model selection: 1,109 S5K frames. | |
| All 1,972 frames from the separate straight/grid-row layout are used exclusively for testing. No straight/grid-row frame is included in training or validation. | |
| This is a simulation-only dataset. The available data do not establish robustness to real plantation conditions, diverse tree geometry, graded occlusion severity, or uneven terrain. | |