Modern Automatic Speech Recognition (ASR) technology has evolved to identify\nthe speech spoken by native speakers of a language very well. However,\nidentification of the speech spoken by non-native speakers continues to be a\nmajor challenge for it. In this work, we first spell out the key requirements\nfor creating a well-curated database of speech samples in non-native accents\nfor training and testing robust ASR systems. We then introduce AccentDB, one\nsuch database that contains samples of 4 Indian-English accents collected by\nus, and a compilation of samples from 4 native-English, and a metropolitan\nIndian-English accent. We also present an analysis on separability of the\ncollected accent data. Further, we present several accent classification models\nand evaluate them thoroughly against human-labelled accent classes. We test the\ngeneralization of our classifier models in a variety of setups of seen and\nunseen data. Finally, we introduce the task of accent neutralization of\nnon-native accents to native accents using autoencoder models with\ntask-specific architectures. Thus, our work aims to aid ASR systems at every\nstage of development with a database for training, classification models for\nfeature augmentation, and neutralization systems for acoustic transformations\nof non-native accents of English.\n