here people can first upload labeled images, and when new images are uploaded, Microsoft Computer Vision returns new labels for those images. Computer Vision can recognize basic features of pictures, such as colors, as well as common objects in the picture. Based on the characters detected, Computer Vision can also classify the images, based on the trained image dataset. To build the training image set, images are selected from our image database and labelled by hand. In our case, we divided the images into three groups, “good”, “bad” and “acceptable”.