You can gather data resources from many sources. You can get images from clip art, scanned photographs, or original computer artwork. Audio can be created, copied, or taken from analog/digital recordings. And tremendous volumes of text exist in multiple formats. There's no shortage of available resources.
The problem is sorting out what you need from what you don't. Using a DBMS to register resources as they're collected can be extremely helpful.
You should design your database to ensure it contains the information about each resource you want to record. Details associated with each record may depend on the type of resource registered. For example, the database record for an audio resource may include whether it's been recorded at 22.05 kHz or at 44.1 kHz. The database record for an image resource might instead include the number of bits used for the image depth.
Another area helped by the use of a DBMS is tracking data resources built from parts of several different resources. This situation will not be uncommon. Your database design should let you register the new resource into the database. For example, say you register an image showing all U.S. presidents. You then use an image editing package to pull out all presidents elected during the twentieth century. You now have a new image to enter in the database. You may want your database to identify the relationship between the original image and its derivative image.