Distributed cosegmentation via submodular optimization on anisotropic diffusion

By Gunhee Kim, Eric P. Xing, Li Fei-Fei, Takeo Kanade. Vision, 2011

312 citations recorded by the source corpus.

The saliency of regions or objects in an image can be significantly boosted if they recur in multiple images. Leveraging this idea, cosegmentation jointly segments common regions from multiple images. In this paper, we propose CoSand, a distributed cosegmentation approach for a highly variable large-scale image collection. The segmentation task is modeled by temperature maximization on anisotropic heat diffusion, of which the temperature maximization with finite K heat sources corresponds to a K-way segmentation that maximizes the segmentation confidence of every pixel in an image. We show that our method takes advantage of a strong theoretic property in that the temperature under linear anisotropic diffusion is a submodular function; therefore, a greedy algorithm guarantees at least a constant factor approximation to the optimal solution for temperature maximization. Our theoretic result is successfully applied to scalable cosegmentation as well as diversity ranking and single-image segmentation. We evaluate CoSand on MSRC and ImageNet datasets, and show its competence both in competitive performance over previous work, and in much superior scalability.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/b6300de4a7d20d3d6d86f691db1beecc2cb9609b

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Enriching Visual Knowledge Bases via Object Discovery and Segmentation (2014, background): “Our work is also related to co-segmentation, where the task is to simultaneously segment visually similar objects from multiple images at the same time [29, 28, 3, 11, 13, 36].”
Unsupervised object discovery and localization in the wild: Part-based matching with bottom-up region proposals (2015, methodology): “On the other hand, cosegmentation [25, 29, 40] and colocalization [12, 27, 51] assume less supervision and only require the image collection to contain a single dominant object class, allowing noisy images to some degree.”
Unsupervised Multi-class Joint Image Segmentation (2014, methodology): “[7] is designed for multi-class segmentation and [8] and [11] are state-of-the-art foreground-background cosegmentation methods.”
Co-Saliency Detection via a Self-Paced Multiple-Instance Learning Framework (2017, background): “…topics mainly lies in three-fold aspects: 1) Co-saliency detection only focuses on discovering the common and salient objects from the given image groups while co-segmentation methods additionally tend to possibly precisely segment out the similar but non-salient background regions [20], [21].”
Clothing Co-parsing by Joint Image Segmentation and Labeling (2014, background): “, more than 50 categories in the Fashionista dataset [30]; the categories are relatively fewer in existing co-segmentation systems [9, 7].”
Semantic Object Selection (2014, methodology): “Cosegmentation methods [9, 8, 10, 19] operate on multiple input images and select in each image a common object.”
ImageNet Auto-Annotation with Segmentation Propagation (2014, background): “Our work is also related to co-segmentation, where the task is to segment multiple images at the same time [14,15, 8,71,34,35,45].”
Salient object detection: A survey (2014, background): “tion. First, co-saliency detection algorithms only focus on detecting the common salient objects while the similar but non-salient background might be also segmented out in co-segmentation approaches [98,158]. Second, some co-segmentation methods, e.g., [16], need user input to guide the segmentatio”
Learning Discriminative Part Detectors for Image Classification and Cosegmentation (2013, unknown): “Cosegmentation [19, 25, 34] is a challenging task in computer vision, since it involves a weak form of supervision, i.”
Multi-class cosegmentation (2012, methodology): “In cosegmentation algorithms, visual and spatial consistency is usually enforced using binary terms based on total variation [4] or the Laplacian of similarity matrices [5, 8].”
On learning to localize objects with minimal supervision (2014, unknown): “While in this paper we do not address pixel-level segmentation, we employ ideas from co-segmentation: the intuition behind our submodular cover framework in Section 4 is shared with CoSand (Kim et al., 2011).”
Beyond pixels: A comprehensive survey from bottom-up to semantic image segmentation and cosegmentation (2015, methodology): “from user-provided bounding boxes which enclose the objectof-interest, and over-segments the images into superpixels by [35].”
Zero-Shot Video Object Segmentation via Attentive Graph Neural Networks (2019, methodology): “Early methods usually formulate IOCS as an energy function defined over the whole or a part of the image set and consider intra- and inter-image cues [64, 25, 52, 65].”
Efficient Image and Video Co-localization with Frank-Wolfe Algorithm (2014, background): “Recent research has studied ways of leveraging this data, such as weakly supervised localization [14], co-segmentation [21, 23, 38], and co-localization [37, 44].”
Segmentation Propagation in ImageNet (2012, unknown): “Our work is also related to co-segmentation, where the task is to segment multiple images at the same time [3, 12, 15, 16, 27].”
Unsupervised Joint Object Discovery and Segmentation in Internet Images (2013, background): “[12] proposed an optimization which scales up to even larger datasets.”
Unsupervised Image Matching and Object Discovery as Optimization (2019, background): “Several alternatives are possible, including weakly supervised approaches that rely on readily available meta-data [2, 9] or image-level labels [14, 23, 24, 25, 39, 45] instead of more complex annotations such as bounding boxes [17, 38] or object masks [20] as supervisory signal; semi supervised met”