Deep neural networks (DNN) have achieved tremendous development in the past decade. While many DNN-driven software applications have been deployed to solve various tasks, however, they could also produce incorrect behaviors and result in massive losses. To reveal the incorrect behaviors and improve the quality of DNN-driven applications, developers often need rich labeled data for the testing and optimization of DNN models. However, in practice, identifying the oracle information of unlabelled data, which describes the expected output for a given input, is often an expensive and time-consuming task. In this paper, we proposed an adaptive test selection method, namely ATS, to alleviate this problem. We incorporate an adaptive selection method into ATS that could select an effective subset from massive unlabelled data. We experiment ATS with four well-designed DNN models and four widely-used datasets in comparison with various kinds of neuron coverage (NC). The results demonstrate that ATS can significantly outperform all test selection methods in assessing both fault detection and model improvement capability of test suites.