PROJECT TITLE :

Extracting 3D Layout From a Single Image Using Global Image Structures - 2015

ABSTRACT:

Extracting the pixel-level 3D layout from a single image is important for various applications, such as object localization, image, and video categorization. Traditionally, the 3D layout is derived by solving a pixel-level classification downside. But, the image-level 3D structure will be very helpful for extracting pixel-level 3D layout since it implies the manner how pixels in the image are organized. During this paper, we propose an approach that initial predicts the global image structure, and then we tend to use the global structure for fine-grained pixel-level 3D layout extraction. In particular, image features are extracted based mostly on multiple layout templates. We then learn a discriminative model for classifying the worldwide layout at the image-level. Using latent variables, we tend to implicitly model the sublevel semantics of the image, which enrich the expressiveness of our model. After the image-level structure is obtained, it's used as the previous data to infer pixel-wise 3D layout. Experiments show that the results of our model outperform the state-of-the-art strategies by 11.7p.c for 3D structure classification. Moreover, we have a tendency to show that employing the 3D structure previous data yields accurate 3D scene layout segmentation.