Sample frames for this project are available under NDA. Write to bilal@wortel.ai and we will share the relevant before and after pairs.
Challenges
- Detecting trees at different distances and sizes.
- Partially visible or occluded trees had to be annotated accurately.
- Overlapping tree canopies made individual trees difficult to isolate.
- Trees whose branches or foliage were closely connected were hard to separate.
- Complex backgrounds contained buildings, roads, crops and other vegetation.
- Varying lighting conditions and camera angles affected tree visibility.
- Accurate bounding boxes around irregular tree shapes had to be maintained.
Related work
Ground Cover Quadrat Segmentation
2,275 cover-crop plot images annotated in Labelbox: plot corners boxed, then every pixel classed as groundcover, weed, residue, crop or soil.
Almond Tree Detection
Bounding boxes around individual almond trees in top-down drone imagery, where shadows and neighbouring vegetation mimicked real trees.
Apple Detection
Bounding boxes around individual apples on tree branches, where dense leaves and occlusion hid much of each fruit.