TALDS: A Transfer-Active Learning-Driven Siamese Network for Bi-temporal Image Classification

A novel Transfer-Active Learning-Driven Siamese network for bi-temporal image classification (TALDS) is proposed. It incorporates transfer learning (TL) and active learning (AL) techniques to facilitate the selection of informative samples in an iterative process. This approach allows the model to learn from different domains efficiently, improving its accuracy and robustness. TALDS network shows promising results for bi-temporal image classification tasks and is therefore a valuable contribution to computer vision. The Siamese network architecture enables the network to learn to extract coherent features from bi-temporal images, allowing for accurate image classification. Experimental results demonstrate the effectiveness of our framework in detecting desertification using satellite images. The proposed method offers numerous possibilities for implementation in different domains, including environmental monitoring and remote sensing.