A Comparison of Affine Region Detectors

By Krystian Mikolajczyk, Tinne Tuytelaars, Cordelia Schmid, Andrew Zisserman, Jiri Matas, Frederik Schaffalitzky, Timor Kadir, Luc Van Gool. 2005

3,519 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

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 Media Technology

Citations in the corpus

Aggregating Local Image Descriptors into Compact Codes (2012, methodology): “The regions of interest are extracted using the Hessian rotation-and afﬁne-invariant region detector [25] and described by the SIFT descriptor [7].”
What Makes for Effective Detection Proposals? (2015, methodology): “In this work we will perform an extensive apples-to-apples comparison of the 12 methods (plus 4 baselines) listed in table 1.”
Cut, Paste and Learn: Surprisingly Easy Synthesis for Instance Detection (2017, methodology): “Early approaches, such as [6], heavily depend on extracting local features such as SIFT [30], SURF [3], MSER [32] and matching them to retrieve instances [29, 48].”
How good are detection proposals, really? (2014, background): “There are some salient aspects of the result curves in ﬁgure 3 that need some explanation.”
AUV Navigation and Localization: A Review (2014, methodology): “3For a review and comparison of some methods, see [123].”
GMS: Grid-Based Motion Statistics for Fast, Ultra-Robust Feature Correspondence (2017, methodology): “Strecha [37]3 and VGG [25] are standard datasets with significant widebaselines and good ground-truth.”
Deformable Medical Image Registration: A Survey (2013, neutral): “[249].”
Key.Net: Keypoint Detection by Handcrafted and Learned CNN Filters (2019, neutral): “Those detectors were further extended to handle multi-scale and affine transformations [6, 19].”
Aggregating local descriptors into a compact image representation (2010, methodology): “More precisely, the regions of interest are extracted using the Hessian affine-invariant region detector [14] and described by the SIFT descriptor [12].”
Visual simultaneous localization and mapping: a survey (2012, background): “Mikolajczyk et al. (2005) made an evaluation of the performance of these algorithms with respect to viewpoint, zoom, rotation, out-of-focus, JPEG compression and lighting changes.”
CZECH TECHNICAL UNIVERSITY IN PRAGUE (2005, methodology): “On the other hand, wide baseline SfM based methods, which use richer features such as MSER [25], Laplacian-Affine, Hessian-Affine [27], SIFT [23], and SURF [2], are capable of producing feasible tentative matches under large changes of visual appearance between images induced by rapid changes of cam”
Performance Evaluation of 3D Keypoint Detectors (2012, methodology): “2010; Chen and Bhanu 2007; Zhong 2009; Novatnack and Nishino 2008; Unnikrishnan and Hebert 2008; Akagunduz and Ulusoy 2007; Zaharescu et al. 2009; Fadaifard and Wolberg 2011; Knopp et al. 2010; Castellani et al. 2008; Sun et al. 2009). Compared to its 2D counterpart, and to 3D descriptors as well, r”
Person Reidentification Using Spatiotemporal Appearance (2006, methodology): “Given an image of a person, the Hessian Affine invariant interest operator [11] is used to nominate points of interest.”
Survey of Pedestrian Detection for Advanced Driver Assistance Systems (2010, methodology): “The idea is to use a keypoints detector, Hessian- Laplace [104] in this case, then compute a shape context descriptor [105] for each keypoint, and finally, cluster them to construct a codebook.”
Description of interest regions with local binary patterns (2009, background): “Since this paper focuses on interest region description, we refer the reader to [6] for more information on interest region detection.”
Illumination Invariant Imaging : Applications in Robust Vision-based Localisation , Mapping and Classification for Autonomous Vehicles (2014, background): “The “Leuven/Light” benchmark dataset for point feature illumination invariance [11] consists only of images taken from the same location with varying exposure, which does not represent true variation in outdoor illumination due to sunlight and shadow.”
Comparative Evaluation of Hand-Crafted and Learned Local Features (2017, neutral): “[28] evaluate affine region detectors by introducing standard metrics and small-scale datasets under various photometric and geometric image transformation.”
Fine-Tuning CNN Image Retrieval with No Human Annotation (2017, methodology): “It uses Hessian affine local features [57], RootSIFT descriptors [58], and a fine vocabulary of 16M visual words [59].”