urbangreenspaces-satellite-imagery

(ONGOING) Developing a Generalizable Computer Vision Model for Detecting Urban Green Spaces and Outdoor Sports Facilities in Aerial Imagery

Current results

Next Steps

• Collect and annotate additional aerial images featuring open grounds, and incorporate some images from the DOTA dataset.
• Train and evaluate various object detection models (EfficientDet, RetinaNet, and SSD), including currently used, on aerial images that include all three classes: parks, open grounds, and specialized sports facilities.
• Compile the results in a table and provide an interpretation of the findings.