About this document

PanoContext: 3D Context Model for Rooms by Vengatesan G is a document available to read on EtoBox.

The document proposes a whole-room 3D context model to understand indoor scenes from single panoramic images. The method generates 3D hypotheses for room layout and object bounding boxes using various image cues. An SVM then ranks these hypotheses by combining bottom-up and top-down context information from a training dataset of annotated panoramas. Experiments show the method can achieve state-of-the-art object detection performance using only 3D context without image classifiers.

Author

Vengatesan G

Language

EN