LC4-DViT: Land-cover Creation for Land-cover Classification with Deformable Vision Transformer
Researchers have developed LC4-DViT, a novel framework for land-cover classification using a deformable Vision Transformer. This approach combines generative data creation with a deformation-aware backbone to improve accuracy and handle geometric distortions in high-resolution imagery. The system synthesizes class-balanced training images using GPT-4o-generated descriptions and achieves state-of-the-art results on benchmark datasets, demonstrating strong transferability and improved attention alignment with relevant structures. AI
IMPACT Introduces a novel approach for high-resolution land-cover mapping by combining generative AI with specialized vision transformers.