In my understanding, domain randomization is one method of diversifying the dataset to achieve a better shot at domain adaptation. Am I wrong?
you are kind of right. but no necessarily. domain randomization : when you widen the range of your training data parameters to make your model more generalized. this can be done for any purpose. (even if you are not doing domain adaptation).
domain adaptation : when you train your model on data from a certain domain (lets say to detect cars) and then test it on data from another domain (lets say for detecting trucks).
- the data you used earlier helped "pre-train" your model.
- you may "fine-tune" this model with whatever small amounts of data you have available from the test domain (trucks in our case).
- you may also use the concepts of domain randomization (make sure you include different varieties of car images) during your training.
these things are optional, and do not always HAVE to go together. Hope this helps!