Markus Vogelbacher from Karlsruhe Institute of Technology created a benchmark dataset for detecting refuse-derived fuel particles in a rotary kiln combustion chamber. It contains 50 annotated 2D images and corresponding 3D point clouds, plus a larger sequence of 2010 images. The dataset includes labeled ground truth images, particle coordinate lists, and Matlab code for visualization.
Use Cases
- Training object detection models for fuel particles based on labeled 2D images.
- Developing 3D particle tracking algorithms based on the provided point cloud data.
- Benchmarking computer vision systems for industrial combustion analysis based on the annotated ground truth.
- Analyzing particle behavior (burning vs. non-burning) in a rotary kiln based on the labeled categories.
Strengths
- Includes 50 annotated 2D images with ground truth labels for five categories.
- Provides corresponding 3D point cloud data for the same 50 images.
- Offers a larger sequence of 2010 images for extended analysis.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count and dataset size are unknown, which may limit suitability assessment.
Provenance
- Source
- Karlsruhe Institute of Technology
- Collection Method
- Likely captured using a light-field camera in an industrial rotary kiln combustion chamber.