<p>The accurate and fast detection of landslides from remote sensing data plays a vital role in disaster management and emergency rescue operations. With the notable advancement in computer vision and remote sensing technology, the manual interpretation/identification of landslides from earth observational platforms has shifted towards an automated technique based on deep learning algorithms. Moreover, the attention mechanism inspired by the human vision system has grown significantly in object recognition problems. The detection of landslides is challenging due to their distinct spatial characteristics and the complex surroundings, leading to inaccurate results. In this regard, the original YOLOv5 network is optimized by the introduction of the attention mechanism for landslide event detection from remote sensing images. Particularly, the convolutional block attention module (CBAM), efficient channel attention, global attention mechanism, and coordinate attention are embedded discretely within YOLOv5 at multiple locations. The dataset used for experimentation is acquired from diverse platforms (satellite and unmanned aerial vehicles) of distinct regions. The evaluation index includes f-score, precision, recall, and mean average precision. The experimental results showed that the YOLOV5n + CBAM (f-score = 0.985, considering satellite images) outperformed other models proposed in this study. Further, the obtained results are compared with the previous studies that applied similar data for research. The quantitative assessment indicates that the proposed work will be an important support for generating landslide susceptibility maps, emergency response, and the development of early prediction models through the identification of potential landslides in large geographical areas. </p>