Aerial perception · Residential environments
Aerial Weed Detection
A4AWD studies aerial weed detection in residential lawns, where weeds can be small, sparse, irregular, and difficult to distinguish from surrounding turfgrass.
01 / Research frame
The problem
Models trained on ground-level agricultural imagery face a substantial shift when the camera moves to an aerial viewpoint. In residential lawns, the target plants may also occupy only a small part of each frame and share color and texture with healthy grass.
The approach
The project develops a weed-specific, inference-time augmentation method for aerial imagery and evaluates it with field data collected across residential properties. The public dataset supports follow-on work and reproducible comparisons.
Contributors
- Jiaxin Du
- Shengqing Xia
- Chunyi Peng
03 / Associated work
Associated publications
A4AWD: Augmenting Aerial Weed Detection Over Residential Lawns
Jiaxin Du, Shengqing Xia, Chunyi Peng · International Conference on Unmanned Aircraft Systems (ICUAS)