Abstract: Intense rainfall commonly triggers spatially clustered landslides, posing serious threats to lives and infrastructure. Rapid and accurate identification of these events is critical for emergency response and risk assessment. Although deep learning has achieved strong performance in remote-sensing-based landslide detection, high-quality annotated datasets tailored to rainfall-induced clustered landslides remain limited, constraining model improvement and generalization assessment. This study constructs a high-resolution remote-sensing labeled dataset for rainfall-induced clustered landslides. The dataset covers three representative clustered landslide events triggered by intense rainfall in central and eastern China in 2024. A unified interpretation protocol was established, and landslides were manually interpreted from 3 m-resolution PlanetScope imagery. The interpretation results were cropped and screened to produce 3,190 image patches of 512 x 512 pixels, each containing an average of approximately 19 landslides. The dataset includes post-event image patches cropped from PlanetScope imagery, manually interpreted landslide label files, and corresponding mask files for deep-learning model training and validation. Validation in representative areas was performed by comparison with UAV-based interpretation results (336 landslides). In the corresponding areas, 365 landslides were identified, including 45 false positives and 16 missed detections. False positives were mainly associated with artificial slopes, exposed bedrock, and gullies, indicating overall high annotation accuracy. The dataset provides an important data foundation for intelligent identification, hazard assessment, and generalization studies of rainfall-induced clustered landslides.