Quick Shift and Kernel Methods for Mode Seeking

By Andrea Vedaldi, Stefano Soatto. 2008

998 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Geometry and Topology

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

In Geometry and Topology

Citations in the corpus

Anchors: High-Precision Model-Agnostic Explanations (2018, methodology): “Image Classification: When explaining the label prediction for an image we follow Ribeiro, Singh, and Guestrin (2016b), segmenting the image into superpixels (Vedaldi and Soatto 2008) and using the presence or absence of these superpixels as the interpretable representation.”
Multiple kernels for object detection (2009, methodology): “In the training images there will be a number of rectangles Ri associated with w; similarly to [17], these are aggregated into a single rectangle using mean-shift clustering (using [23] for speed).”
Superpixel tracking (2011, background): “In particular, superpixels have been one of the most promising representations with demonstrated success in image segmentation and object recognition [18] [15] [22] [12] [17].”
Superpixels: An evaluation of the state-of-the-art (2016, background): “Reference (Google Scholar Citations) Color Vedaldi and Soatto [62], 2002 (376)”
CNN-Enhanced Graph Convolutional Network With Pixel- and Superpixel-Level Feature Fusion for Hyperspectral Image Classification (2020, methodology): “In this section, we test several state-of-the-art superpixel segmentation methods on the CEGCN, including the SLIC [45], the quick shift (QS) [52], the Felzenszwalb and Huttenlocher’s (FH) method [53], the linear spectral clustering (LSC) [54], and the SEEDS [55].”
SLIC Superpixels Compared to State-of-the-Art Superpixel Methods (2012, methodology): “We therefore performed an empirical comparison of ﬁve state-of-the-art superpixel methods [8], [23], [26], [25], [15], evaluating their speed, ability to adhere to image boundaries, and impact on segmentation performance.”
SLIC Superpixels ? (2010, background): “Quick-shift [10] (QS08) is also a mode-seeking segmentation scheme like Mean-shift, but is faster in practice.”
Robust Superpixel Tracking (2014, methodology): “In particular, superpixels have been applied to image segmentation and object recognition with demonstrated success [20], [23], [24], [27], [28].”
Salient object detection: A survey (2014, neutral): “Quick-Shift [169], as a variant of the Mean-Shift algorithm, simply moves each point in the feature space to”
SEEDS: Superpixels Extracted Via Energy-Driven Sampling (2012, methodology): “Other methods not based on graphs, are Watersheds [16], Turbopixels [15], a Geodesic distances-based approach [17], SLIC [12], Consistent Segmentation [18] and Quick Shift [19].”
Segmentation-Based Image Copy-Move Forgery Detection Scheme (2015, methodology): “Among them the methods in [32] and [34] are more favorable owing to their comparatively lower complexity.”
Harmony Potentials (2011, methodology): “Regions are created by over-segmenting the image with the quick-shift al-gorithm (Vedaldi and Soatto, 2008) using the same parameters as Fulkerson et al (2009).”
Multi-scale object candidates for generic object tracking in street scenes (2016, methodology): “We iteratively smooth the density map and identify blobs (clusters) around modes in the density map using Quick-Shift [31] at each scale.”
Ontology-Based Semantic Image Segmentation Using Mixture Models and Multiple CRFs (2016, methodology): “Several unsupervised segmentation algorithms have hence been used for higher level region-based feature representation such as JSEG [4], mean shift [5], graph-based segmentation [6], [7], quick shift [8], TurboPixel [9] and SLIC superpixel [10].”
Towards Automatic Concept-based Explanations (2019, methodology): “The segmentation step is performed using SLIC [2] due to its speed and performance (after examining several super-pixel methods [10, 26, 41]) with three resolutions of 15, 50, and 80 segments for each image.”
Semi-Supervised Kernel Mean Shift Clustering (2014, result): “We also note that a similar derivation for fixed bandwidth mean shift algorithm was given in [39].”