The paper presents an end-to-end trainable network for segmenting farmlands with contour levees from high-resolution aerial imagery. The network uses a fusion block that includes multiple voting blocks for image segmentation and classification. The method achieved an average accuracy of 94.34%, improving the F1 score by 6.96% and 2.63% compared to other methods. The network assigns the most likely class label of a segment to its pixels, learning the concept of farmlands rather than analyzing pixels separately.
Publication date: 29 Sep 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2309.16561
