Remove furniture
Images of different kinds of places in a building is given, the place is usually furnished with furnitures, books, lightings etc. I am to coat any of the objects I wish to remove from the image with black ink and prompt the AI to "remove the chandelier" - for example I will do this for three different objects and then reverse engineer the initial prompts to see if it will achieve the initial setup of the image, then I judge the performance on percentages