3260 papers • 126 benchmarks • 313 datasets
This task has no description! Would you like to contribute one?
(Image credit: Papersgraph)
These leaderboards are used to track progress in junction-detection-10
No benchmarks available.
Use these libraries to find junction-detection-10 models and implementations
No datasets available.
No subtasks available.
A learning-based approach to the task of automatically extracting a "wireframe" representation for images of cluttered man-made environments and two convolutional neural networks that are suitable for extracting junctions and lines with large spatial support are proposed.
A bottom-up model for simultaneously finding many boundary elements in an image, including contours, corners and junctions, which allows it to succeed at high noise levels, where other methods for segmentation and boundary detection fail.
Adding a benchmark result helps the community track progress.