We are temporarily pausing the project due to unforeseen circumstances, and will re-launch as soon as possible.
Also, this project recently migrated onto Zooniverse’s new architecture. For details, see here.
FAQ
What is this project about?
We’re building a high-quality training dataset to teach computer vision models how to detect and identify urban trees from panoramic street-view photos. Your help marking trees and identifying their species will directly support this AI development.
What kind of photos am I looking at?
Most of the photos are panoramic street-view images (e.g. from vehicle-mounted cameras). These give us a 360-degree view of urban environments. Some photos may be distorted or stitched together, which is normal for panoramic imagery.
I’m not a tree expert. can I still help?
Absolutely! We provide visual guides, tutorials and simple species options. Even if you’re unsure, your input still helps. Multiple volunteers will see each image, so every contribution adds value.
How will my contributions be used?
Your labels will help train and validate machine learning models that detect and classify trees in urban photos. These models can assist in city planning, environmental studies, and real-time biodiversity monitoring.
What happens if I make a mistake?
Don’t worry! Each photo is shown to multiple users, and AI models are trained to work with uncertainty. If you’re not sure, it’s perfectly fine to skip or choose "Unknown".
Will I learn how to identify trees?
Yes! Over time, with help from our guides and repeated participation, you may begin to recognize common tree shapes, bark patterns, and leaf silhouettes, even from street-level views.
What kind of data is being collected?
Each annotation includes:
- Tree location (bounding box),
- User-labeled species (if provided),
- Image metadata (e.g., GPS location),
This data is exported in structured formats for training machine learning models.
What kind of AI or machine learning is this project supporting?
We’re training computer vision models, specifically object detection and image classification algorithms to identify and label urban trees in street-level images.
Will the model generalise to other cities or regions?
That’s our goal, but generalisation depends on training diversity. By collecting images from different urban environments, tree types, and lighting conditions, we can train models that perform well in many regions.