916 hand-annotated Arabidopsis thaliana plants across 56 images provide leaf-level ground truth for computer vision analysis. The dataset includes visible light, top-down timelapse images taken every 15 minutes from a robotic greenhouse system, alongside harvested plant data and supporting software. It was created by Jonathan Bell for the EPSRC-funded project "Dynamic Modelling of Plant Growth with Computer Vision" (grant EP/LO17253/1).
Use Cases
- Training leaf segmentation models based on hand-annotated leaf-level ground truth.
- Developing timelapse analysis algorithms for plant growth using 15-minute interval images.
- Validating computer vision techniques for plant science with destructive measurement data.
- Benchmarking image analysis software for Arabidopsis thaliana studies.
Strengths
- Includes 56 images with 916 hand-marked individual plants for precise ground truth.
- Contains original timelapse images, destructive measurements, and supporting software for analysis.
- Data was acquired using a controlled robotic greenhouse system, suggesting consistent capture conditions.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count and total image count are unknown, which may limit suitability assessment.
- Last update date is unknown; freshness unverified.
Provenance
- Source
- Jonathan Bell, associated with the EPSRC project "Dynamic Modelling of Plant Growth with Computer Vision".
- Collection Method
- Images captured via a robotic greenhouse system in a 15-minute timelapse sequence; plants were periodically sacrificed for destructive measurements.