This tool is an easier interface to the Clarifai API for people who either don't want to code or don't want to bother to learn how to use the API. What this API does in short is that it takes a list of images (in the form of urls) and outputs what an algorithm, also called a model, "sees" in these images.
Different models are specialized in "seeing" different things. Some were trained to pick out colors, others are great at identifying celebrities, so choose the model that most suits you for the job. Remember that they are not perfect, just like humans, so they may make mistakes (sometimes big ones).
The input should be a .csv file with all the image urls. There must be one image per line and a column header at the beginning of the file.