Pixel Objectness

TL;DR

This paper introduces a deep learning framework that generates pixel-level object segmentations for novel images, even for unseen categories, by combining weak and strong supervision, improving state-of-the-art results.

Contribution

It presents an end-to-end fully convolutional network that learns to produce foreground masks using limited boundary annotations and image-level labels, enabling generalization to unseen object categories.

Findings

Significantly outperforms previous methods on ImageNet and MIT datasets.

Successfully generalizes to over 1 million images with unseen categories.

Enhances image retrieval and retargeting tasks with high-quality foreground maps.

Abstract

We propose an end-to-end learning framework for generating foreground object segmentations. Given a single novel image, our approach produces pixel-level masks for all "object-like" regions---even for object categories never seen during training. We formulate the task as a structured prediction problem of assigning foreground/background labels to all pixels, implemented using a deep fully convolutional network. Key to our idea is training with a mix of image-level object category examples together with relatively few images with boundary-level annotations. Our method substantially improves the state-of-the-art on foreground segmentation for ImageNet and MIT Object Discovery datasets. Furthermore, on over 1 million images, we show that it generalizes well to segment object categories unseen in the foreground maps used for training. Finally, we demonstrate how our approach benefits image retrieval and image retargeting, both of which flourish when given our high-quality foreground maps.