Researchers have developed DDF2Pol, a novel dual-domain convolutional neural network designed for classifying PolSAR images. This network utilizes parallel real-valued and complex-valued streams to extract complementary information, enhanced by depth-wise convolution and a coordinate attention mechanism. Experiments show DDF2Pol achieves high accuracy on benchmark datasets, reaching 98.16% on Flevoland and 96.12% on San Francisco, with a low parameter count of 91,371. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
RANK_REASON This is a research paper detailing a new model for image classification.