Personal identity documents (IDs) are used for identity verification by many services in the real world. Recent research includes identity document verification using only an image of the document obtained from a mobile phone camera allowing verification procedures to be automatic and remote. Literature shows promising results for verification of standard IDs although the problem of verifying non-standard IDs such as student identity cards remains less explored. Identity verification pipeline often requires information matching , a non-trivial problem in case of non-standard identity cards, lacking in other literature. In this study, we describe a complete system utilizing deep learning methods to verify student identity cards from images captured using a mobile phone camera. Other contributions of this study include a method to correct skew in images using a convolutional neural network, use of state-of-the-art models for text extraction, comparison between algorithms applicable to various stages of the system and a customized fuzzy string-matching algorithm for information verification. This system evaluates an ID card in 2.31s and achieves an F1 score of 0.90054, G-Mean of 0.89442, and average precision of 0.95 in detecting improper ID cards. The system achieves superior performance in verifying student IDs, a type of non-standard IDs. The study also highlights several enhancements and open problems for improving verification.