Current browse context:
cs.GR
Change to browse by:
References & Citations
Computer Science > Computer Vision and Pattern Recognition
Title: BlenderProc
(Submitted on 25 Oct 2019)
Abstract: BlenderProc is a modular procedural pipeline, which helps in generating real looking images for the training of convolutional neural networks. These can be used in a variety of use cases including segmentation, depth, normal and pose estimation and many others. A key feature of our extension of blender is the simple to use modular pipeline, which was designed to be easily extendable. By offering standard modules, which cover a variety of scenarios, we provide a starting point on which new modules can be created.
Submission history
From: Maximilian Denninger [view email][v1] Fri, 25 Oct 2019 16:53:12 GMT (3307kb,D)
Link back to: arXiv, form interface, contact.