See the data
behind the images.
Synthetic plant imagery with matching detection boxes and visible plant masks. Open an example, inspect its labels, and try the files in your own pipeline.
Free for internal research and evaluation, including within companies. Sample use terms
AMBER / GENERATED LABELS · COTTON IS UNLABELED CONTEXT
Full canopy
An early-vegetative Palmer plant with cotton context, viewed from above.
Choose an example to see different plant stages, target sizes, and visibility.
80 Palmer examples
20 negative examples
JPEG images
Lossless PNG masks
Five related views each
Three plant growth stages
Ready to inspect.
- COCO boxes and visible instance masks.
- Matching detection text files and semantic masks.
- Camera settings and layout groups for each image.
- A guide and a Python integrity-check script.
A focused starting point.
This sample uses one overhead camera setup and a fixed daylight profile. Palmer is the labeled target; cotton provides context. The labels come from the generated scene.
It demonstrates the imagery and annotation format. Field performance needs a separate evaluation on real images from the task you care about.
Image quality and evaluation notes
The download contains 640 × 640 JPEGs converted from the original renders at quality 92 with full chroma resolution. Masks and labels are unchanged. Browser previews are compressed separately.
The sample is drawn from an existing synthetic export. Related views share layout groups; keep them together when splitting data. No train/test benchmark or detector weights are included.
What does your camera see?
Tell us the crop, camera setup, and examples your team needs.