The goal of this project is to create a database of labeled image groups and a 'ground truth' mask for each image. First, we create categories of images by collecting images with different objects. Then we segment the target objects in each image category to create the database. Since the process requires considerable image segmentation, we needs a tool that can easily segment out the object we are interested in a given image. This task is sometimes called "ground truthing", which is used to be done by labeling pixel by pixel manually from an image. We are seeking for an efficient way carried out by programming. The process begins with generating superpixels. A superpixel is a group of adjacent pixels within an image which contain similar properties. By dividing an image into superpixels, we can work on these larger units rather than to look on each pixel. After an image is segmented into superpixels, a GUI tool is developed to select these superpixels related to our target objects. The result will be an image 'mask', which keeps our interested objects and covers the rest region.