TLSTMF-YOLO: Transfer Learning and Feature Fusion Network for Earthquake-Induced Landslide Detection in Remote Sensing Images
Citation
Zhang, F., Zhu, Q., Wang, Y., Liu, J., Chen, J., Li, W. (2025). TLSTMF-YOLO: Transfer Learning and Feature Fusion Network for Earthquake-Induced Landslide Detection in Remote Sensing Images. IEEE Transactions on Geoscience and Remote Sensing, 63. Link to paper
Abstract
This research combines C3-Swin-Transformer and multiscale feature fusion techniques to enhance detection accuracy and efficiency for earthquake-induced landslides in remote sensing images. The model utilizes a feature extraction layer and Swin-Transformer structure to capture dependencies and preserve spatial information, incorporating Convolutional Block Attention Module (CBAM) to enhance feature representation. A Bidirectional Feature Pyramid Network (BiFPN) optimizes bidirectional cross-scale feature fusion, improving landslide detection accuracy across scales. Testing on Jiuzhaigou dataset achieved precision of 95.7%, recall of 89.9%, and mAP@0.5 of 90.5%. On Luding dataset, the model achieved precision of 96.0%, recall of 90.9%, and mAP@0.5 of 94.5%. Frame processing times are 6.61 ms and 12.2 ms on the respective datasets, demonstrating superior efficiency.