
GeoDiff-SR
4x diffusion super-resolution for agricultural drone imagery
GeoDiff-SR is a conditional denoising-diffusion model that enhances the resolution of agricultural UAV images by a factor of four while keeping reconstructions faithful to the underlying scene. It recovers fine canopy detail without the hallucinated texture typical of GAN-based super-resolvers, so agronomic measurements stay reliable.
The model was trained on 2,136 sweet corn and green bean images collected over three growing seasons (2020 to 2023) at the UF/IFAS Tropical Research and Education Center. Try it on bundled low-resolution plot frames in your browser.
- Mean opinion score 0.98, vs 0.82 for Real-ESRGAN and 0.45 for bicubic
- Canopy-coverage error 0.71%, vs 2.37% for Real-ESRGAN
- Trained on 2,136 sweet corn and green bean UAV images (2020-2023)
- Rated in a blind human-perception study with 38 participants






