A Perfectly Accurate, Synthetic dataset featuring a virtual railway EnVironment for Multi-View Stereopsis (RailEnV-PASMVS) is presented, consisting of 40 scenes and 79,800 renderings together with ground truth depth maps, extrinsic and intrinsic camera parameters, pseudo-geolocation metadata and binary segmentation masks of all the track components. Every scene is rendered from a set of 3 cameras, each positioned relative to the track for optimal 3D reconstruction of the rail profile. The set of cameras is translated across the 100 m length of tangent (straight) track to yield a total of 1995 camera views. Photorealistic lighting of each of the 40 scenes is achieved with the implementation of high-definition, high dynamic range (HDR) environmental textures. Additional variation is introduced in the form of camera focal lengths, camera location and rotation parameters and shader modifications for materials. Representative track geometry provides random and unique vertical alignment data for the rail profile for every scene. This primary, synthetic dataset is augmented by a smaller photograph collection consisting of 320 annotated photographs for improved semantic segmentation performance. The combination of diffuse and specular properties increases the ambiguity and complexity of the data distribution. RailEnV-PASMVS represents an application specific dataset for railway engineering, against the backdrop of existing datasets available in the field of computer vision, providing the precision required for novel research applications in the field of transportation engineering. The novelty of the RailEnV-PASMVS dataset is demonstrated with two use cases, resolving shortcomings of the existing PASMVS dataset.
Keywords: Blender; ECEF; Geolocation; Ground truth depth maps; Multi-view stereopsis; Railway engineering; Semantic segmentation; Synthetic data.
© 2021 The Author(s). Published by Elsevier Inc.