Storing docker images that aren't needed costs money unnecessarily. Not storing images you still need can cause outages. ECR (and other equivalents) can be configured to delete images not pulled for some period of time, but if a pod is left running for longer than that then its image may be deleted, leading to breakage if e.g. the node hosting the pod then goes down.
This tool examines all of the pods on a cluster, and pulls the metadata for each image used. If run on a regular schedule, this should prevent images from being cleaned up while they are still in use.