SCNet: Lightweight Spatial–Channel Attention Network for Remote Sensing Change Captioning

Change captioning has become essential for accurately describing changes in multi-temporal remote sensing data, providing an intuitive way to monitor Earth’s dynamics through natural language. However, existing change captioning methods face two key challenges. First, the use of multistage feature fusion strategies aims to achieve better change detection results by combining information from multiple stages. This approach leads to high computational demands, as it requires intensive processing across different feature levels. Second, previous methods mainly focus on feature extraction and fusion along the image spatial dimensions, overlooking the importance of channel-wise information. This leads to insufficient semantic extraction, resulting in inadequate detail in object descriptions and limiting the ability to fully capture the changes. To solve these challenges, we propose Spatial-Channel Attention Encoder based on the transformers model for remote sensing change captioning, which named SCNet. In particular, SCNet integrates a Spatial-Channel Attention Encoder, a Difference-Guided Fusion module, and a Caption Decoder. Compared to typical models that require multi-stage fusion in transformer encoder and fusion module. By jointly modeling spatial and channel information in Spatial-Channel Attention Encoder, our approach significantly enhances the model’s ability to extract semantic information from objects in multi-temporal remote sensing images. Extensive experiments validate the effectiveness of SCNet, achieving CIDEr scores of 140.23% on the LEVIR-CC dataset and 97.74% on the DUBAI-CC dataset, surpassing current state-of-the-art methods. The code and pre-trained models will be available at SCNet.