Yes, you can deal with unstructured data, and you can label images. For datasets with images, we assume are either saved in a folder, and their paths captured in a csv or excel file. Then to train the dataset in DataRobot, we assume that you compressed the csv and the folder with images to a zip file. You can find more info here For text, we automatically detect text, and create on the fly blueprints that can pre-process it. Furthermore, we provide, in composable ML (At the blueprint level) functions that can further extend text manipulation, like sentiment detection, SPACY or text embedding, etc. You can find more info here We also can deal with geo-locations. So, if you have longitude and latitude of a location, we will provide you with a geo map. You can find detail info here.
As my colleague @dalilaB has mentioned our DataRobot AutoML platform can ingest and model unstructured data in the various formats listed.
Our DataRobot DataPrep tool is more designed for structured data, but as @dalilaB suggested, you can refer to the zipped images in a column referencing a relative directory path where they are kept in a csv or excel file.
We have courses in DataRobot University that cover the DataRobot Data Prep tool as well as Visual AI modeling and the formats you can use listed here: