Introduced in One-Shot Segmentation in Clutter
Dataset for one-shot segmentation.
Source: One-Shot Segmentation in Clutter
An improved model that attends to multiple candidate locations, generates segmentation proposals to mask out background clutter and selects among the segmented objects is introduced, suggesting that such image recognition models based on an iterative refinement of object detection and foreground segmentation may provide a way to deal with highly cluttered scenes.
