Landslide information in the form of inventory is crucial for landslide risk assessment and mitigation. To prepare a landslide inventory, cognitive interpretations are often considered by the researchers. Although, it is a time-consuming process and requires specific expertise to generate a reliable inventory, in this study, a deep learning (DL) model is configured with the aim of reducing human supervision for inventory preparation. Here, a ResU-Net model is used with Sentinel-2 satellite RGB image and applied for the parts of Malappuram district of Kerala, India. For model training and testing, recent landslides triggered during 2018–2019 monsoon period are used in terms of training and testing datasets. For model training, the number of patches of size 128 × 128 is generated along with their labelled images and accordingly used for model training. To gauge the model performance, precision, recall, F1-score, and loss function are evaluated. The accuracy assessment shows promising results for this area. Therefore, it is envisaged that the developed model can be applied to other landslide-affected areas, where the landslide inventory information is scarce in space and time. A training dataset with 126 samples achieved a loss of 0.0489, a precision value of 0.98, recall of 0.93 and F1 score of 0.95. For testing dataset, results obtained were loss = 0.21, precision = 0.95, recall = 0.79 and F1 score = 0.78. The assessment carried out in the proposed study reveals that deep learning models with considered datasets for training are very effective for detecting event based landslides.