Loading...

Room Layout Estimation and Perception through Cross-task Consistency and Knowledge Fusion

Saberi, Ali | 2021

583 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54379 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Bagheri Shouraki, Saeed
  7. Abstract:
  8. Wall detection, as a part of scene understanding and indoor 3D modeling, can be used in robotic, architecture, and augmented reality. In robotic, scene analysis and understanding is knows as one of the main steps in robot navigation and simultaneous localization and mapping and walls need to be detected to form a 3D map of indoor environment. Ceiling and floor are also important elements of an indoor environment, therefore we can see our problem as a room layout estimation if we consider ceiling and floor. Using a deep learning structure, we estimate room layout in a semantic segmentation manner. Our approach is real-time. Our proposed method uses a deep fully-convolutional network, layout degeneration for data augmentation, and consistency constraints and smoothness term in loss function. By using cross-task consistency, adaptive edge penalty, and smoothness term, unlike most of deep learning-based room layout estimation methods, our method is able to estimate room layout without any post-processing steps (e.g. optimization, and ranking layout proposals). Furthermore, since we use semantic segmentation, we can estimate non-cuboid layouts even when we train our network using a cuboid dataset. Moreover, we proposed a method for door detection in indoor environment. Our method uses line segments and room layout information to effectively detect doors and distinguish between doors and door-like objects. Finally, we evaluate our method on a collected dataset of doors
  9. Keywords:
  10. Deep Learning ; Semantic Segmentation ; Scene Detection ; Room Layout Estimation ; Wall Detection ; Door Detection ; Layout Estimation ; Task-Cross Consistency

 Digital Object List