Do deep convolutional neural networks really need to be deep when applied for remote scene classification?

Unsupervised remote sensing image classification with differentiable feature clustering by coupled transformer

“…In the last decade, semantic segmentation methods based on deep convolutional neural networks (CNNs) have demonstrated outstanding results on various benchmark datasets 18 – 21 Recent supervised image classification algorithms grounded in deep learning have exhibited significant enhancements in accuracy 22 , 23 . Semantic segmentation networks featuring a symmetric encoder-decoder structure, such as UNet 24 and its improved versions, 25 are commonly employed.…”