This study used Deep Convolutional Generative Adversarial Networks (DCGANs) to generate images of semi-transparent drinking glasses and of indoor household and office scenes containing drinking glasses. Our training data for the indoor scenes generation consisted of 9200 images where semi-transparent drinking glasses were placed in various locations inside home and office environments. For the generation of semi-transparent drinking glasses, we used 2600 cropped images of drinking glasses. The indoor scenes generated by DCGANs were abstract in nature, capturing the style of the setup but mostly failed to render the drinking glasses properly. For the cropped drinking glasses dataset, the generated images perfectly highlighted the features of the different types of drinking glasses. Finally, we trained several convolutional networks from scratch as classifiers on the DCGANs generated synthetic images of drinking glasses. When tested on the real-world images Vgg_16 performed best among all trained networks with an accuracy of 0.85.