CutMix-CD: Advancing Semi-Supervised Change Detection via Mixed Sample Consistency

Change detection (CD) is an important task in Earth observation. In the past few years, significant progress has been made in supervised CD research; however, change labels are extremely expensive. The semi-supervised CD has attracted increasing attention. In semi-supervised CD, the problem of scarcity of positive samples is magnified. The imbalance of change types (e.g., disappearance and appearance), moreover, exacerbates the missing detection phenomenon. To address the above problems, we propose a semi-supervised CD method: CutMix-CD, which incorporates the change-aware CutMix augmentation into the consistency framework of CD. The semi-supervised learning framework enriches change contexts and places special emphasis on the comparative process, facilitating more robust representations of changes with improved generalization capabilities. First, mixed samples are synthesized using the change-aware CutMix operation. Then, we developed a student path and a teacher path to predict the changes in the original samples and mixed samples, respectively. Finally, the consistency loss is conducted between the two predictions to help the model learn the change information of unlabeled samples. In addition, an unsupervised feature constraint loss is proposed to further optimize the change features. Experiments on four datasets validate the effectiveness of CutMix-CD. It can effectively alleviate the overfitting problem for unbalanced types of changes and even outperforms the fully supervised methods for some challenging samples. The code will be released in https://github.com/SQD1/CutMixCD.