Graphs are pervasive in many applications in which interconnected data are used to represent, explore and predict digital and real-world phenomena. Oftentimes, graph data comes without a predefined structure and in a constraint-less fashion, thus leading to inconsistency and poor quality. In this paper, we present a novel end-to-end schema inference method for property graph schemas that tackles complex and nested property values, multilabeled nodes and node hierarchies. Our method consists of three main steps, the first of which builds upon Cypher queries to extract the node and edge serialization of a property graph. The second step builds over a MapReduce type inference system, working on the serialized output obtained during the first step. The third step analyzes subtypes and supertypes to infer node hierarchies. We present our schema inference pipeline under two variants, namely a label-and a property-oriented variant. Finally, we experimentally evaluate and compare its scalability and accuracy on several real-life datasets. To the best of our knowledge, our work is the first to address schema inference for property graphs.