Abstract:
We detect semantic lines in natural scenes. Many methods treat semantic line detection as a special case of object detection. These methods neglect line characteristics and perform poorly. Lines are simpler geometrically than complex objects and can be parameterized with fewer arguments. This paper incorporates the classical Hough transform technique into deeply learned representations and proposes a one-shot end-to-end line detection framework to better exploit lines. Hough transforms deep representations into parametric domains for line detection by parameterizing lines with slopes and biases. We aggregate features along candidate lines on the feature map plane and assign them to parametric domain locations. Thus, detecting semantic lines in the spatial domain becomes detecting points in the parametric domain, making post-processing, such as non-maximal suppression, more efficient. Our method also simplifies contextual line feature extraction for accurate line detection. We also create a large line detection dataset and an evaluation metric to assess line detection quality. Our method outperforms state-of-the-art alternatives on our proposed dataset and another public dataset.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
