The Internet has become a big thing nowadays. The current growth of infrastructure allows more people to reach the Internet. The Internet has dramatically influenced the e-commerce market. E-commerce has grown significantly, and many new e-commerce sites, including Instagram with their Instagram Shopping, have started appearing. In Instagram Shopping, there is no label for the product, which is essential in e-commerce. This project is about labeling the fashion image using semantic segmentation. The method that was used is U-Net. In this project, we trained the model with 2 different epochs, 10 and 50. The mean IoU on 10 epoch model is 0.044, and 50 epoch model, is 0.092. Moreover, the weighted F1-Score on 10 epoch model is 0.760 and 50 epoch model 0.819. Subsequently, the label output result was filtered with a pixel-count threshold to reduce the noisy label on the output.