Common (non-text) datasets to test classification algorithms?

Question

Classification

I have some classifier (a variation on the theme of the SVM classifier, if you like). I want to evaluate its accuracy, as well as how well it estimates the probability that such and such a point belongs to such and such a class.

I have to ask: what data sets are considered the most common for evaluating classifiers? If you like, the most textbook?

I am especially interested in non-text data, because I have a lot of text corpora anyway, but I would like to check on something else.

Answer 1

Inria for example.
Daimler pedestrian db

Answer 2

If something classic - Iris dataset - classification of iris flowers according to a small set of parameters. But this set is more of a toy set - there are too few samples.
More seriously, then MNIST digits is a classification of handwritten digits.