Indian Multi-Terrain Segmentation Dataset (IMTSD) is a dataset for satellite imagery-based multi-class segmentation in coastal regions of India. The dataset comprises 500 tiles of 256x256 pixel resolution, collected from Sentinel-2 multispectral imagery. Each tile is provided in three different forms of imagery: true RGB, false color, and natural color, making a total of 1500 images. The dataset includes a wide variety of terrain types such as urban areas, vegetation, water bodies, roadways, buildings, and bare land. The dataset is preprocessed using an optimized method for better visualization of the features by applying percentile clipping, histogram stretching, and gamma correction. Indian Multi-Terrain Segmentation Dataset is primarily meant for segmentation and land use/cover classification purposes.