In this study, landslide risk assessment was conducted in the Renhe District, Panzhihua City, China. Firstly, based on 190 landslide points and 10 influencing factors, the landslide hazard was assessed using three models: random forest (RF), eXtreme Gradient Boosting (XGBoost), and Tabular Prior-data Fitted Network (TabPFN). The results indicate that the RF and XGBoost models exhibit comparable performance, both demonstrating strong generalization and accuracy, with the RF model achieving superior generalization, as evidenced by an area-under-the-curve (AUC) value of 0.9471. While the AUC value of TabPFN is 0.9243, indicating higher accuracy, it also poses a risk of overfitting and is therefore more suitable for applications involving small sample sizes and the need for rapid responses. The vulnerability assessment utilized the Analytic Hierarchy Process (AHP) to determine the weights of four disaster-bearing bodies, with sensitivity analysis revealing that road type was the most sensitive vulnerability factor. Finally, the landslide risk-assessment map of the Renhe District was produced by integrating the landslide hazard assessment map with the vulnerability assessment map. The findings indicate that the high-risk zones comprised 2.08% of the research region, which includes three principal train stations and necessitates enhanced protective measures. The medium-risk zones comprise 34.23% of the total area and are scattered throughout the region. It is important to enhance local capabilities for landslide monitoring and early warning systems. Relevant conclusions can provide a significant reference for landslide disaster prevention and mitigation work in the Renhe District and help ensure the safe operation of public transport infrastructure, such as railway stations and airports in the district.

Landslides are frequent and destructive geological disasters. Accurate landslide identification is essential for post-disaster reconstruction and preventing secondary disasters. Deep learning has shown considerable potential for recognizing landslide objects from remote-sensing images; however, existing models still suffer from insufficient detection accuracy in scenarios with complex backgrounds, blurred boundary localization, sample-class imbalance, and difficulty in balancing detection speed and segmentation accuracy. To address these issues, this study investigates landslide identification in the Great Bend of the Yarlung Zangbo River region using improved deep learning models and heterogeneous optical remote-sensing imagery. (1) By introducing the convolutional block attention module (CBAM) into YOLOv8, 82.79% precision was achieved, and the recall improved by 9.56% compared to the original model, reaching a mean average precision of 75.27% while maintaining computational efficiency, outperforming YOLOv5 and the original YOLOv8. (2) Replacing the cross-entropy loss with Focal Loss in DeepLabV3+ improved the landslide edge segmentation by dynamically adjusting the weights of difficult and easy samples. Compared with the original DeepLabV3+, the precision and recall of the DeepLabV3+-FL semantic-segmentation model were improved by 0.26% and 1.93%, respectively, with the mean pixel accuracy and mean intersection over union reaching 81.76% and 62.59%, respectively. Overall, the two improved models enhanced the accuracy of landslide identification and resistance to interference, demonstrating potential for landslide monitoring and emergency response.