Semantic Terrain Points Labeling - Synthetic 3D (STPLS3D)
News:
-
03/31/2023: We are co-organizing the 3rd Urban3D workshop@ICCV2023 with SensatUrban! Please check the Instance segmentation challenge (Codalab) to participate and win the cash award!
-
10/13/2022: Our paper has been accepted as an oral presentation at BMVC2022!
-
03/25/2022: We are co-organizing the 2nd Urban3D workshop@ECCV2022 with SensatUrban! Please check the Instance segmentation challenge (Codalab) to participate and win the cash award!
Our Focus
-
Our project aims to provide a large database of annotated ground truth point clouds reconstructed using aerial photogrammetry.
-
Our database can be used for training and validating 3D semantic and instance segmentation algorithms.
-
We are developing a synthetic data generation pipeline to create synthetic training data that can augment or even replace real-world training data.
​
Our Method
Our designed synthetic data generation pipeline takes full advantage of the off-the-shelf 3D scene generation engine, autonomous vehicles and UAV simulator, and photogrammetry software. Please refer to our paper under the publications tab for details of our designed pipeline and experiment results.
We are providing
-
Three versions of the synthetic data sets that cover approximately 16 sq. Km area with a variety of landscapes (i.e., various buildings styles, types of vegetation, and urban density). Both semantic and instance annotations are available.
-
Four annotated real-world data sets, including USC Campus, Wrigley Marine Science Center, Orlando Convention Center, and a residential area with semantic annotation of selected categories.
-
The annodated 3D point clouds can be downloaded under the download tab. Other data such as annotated source images and 3D meshes are also available upon request.