Segment Augmentation and Differentiable Ranking for Logo Retrieval

Download

index.pdf

Date

2022-1-01

Author

YAVUZ, FEYZA
Kalkan, Sinan

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

138
views

42
downloads

Logo retrieval is a challenging problem since the definition of similarity is more subjective than image retrieval, and the set of known similarities is very scarce. In this paper, to tackle this challenge, we propose a simple but effective segment-based augmentation strategy to introduce artificially similar logos for training deep networks for logo retrieval. In this novel augmentation strategy, we first find segments in a logo and apply transformations such as rotation, scaling, and color change, on the segments, unlike the conventional strategies that perform augmentation at the image level. Moreover, we evaluate suitability of using ranking-based losses (namely Smooth-AP) for learning similarity for logo retrieval. On the METU and the LLD datasets, we show that (i) our segment-based augmentation strategy improves retrieval performance compared to the baseline model or image-level augmentation strategies, and (ii) Smooth-AP indeed performs better than conventional losses for logo retrieval.

URI

https://hdl.handle.net/11511/101848

DOI

https://doi.org/10.1109/icpr56361.2022.9956193

Conference Name

26th International Conference on Pattern Recognition, ICPR 2022

Collections

Department of Computer Engineering, Conference / Seminar

Suggestions

OpenMETU
Core

Context-sensitive matching of two shapes Başeski, Emre; Tarı, Zehra Sibel; Department of Computer Engineering (2006) The similarity between two shapes is typically calculated by measuring how well the properties and the spatial organization of the primitives forming the shapes agree. But, when this calculations are done independent from the context, i.e. the whole set of shapes in the experiments, a priori significance to the primitives is assigned, which may cause problematic similarity measures. A possible way of using context information in similarity measure between shape A and shape B is using the category informatio...
Segmentation based distortion equalizing image coding İmre, Evren; Severcan, Mete; Department of Electrical and Electronics Engineering (2002) In image coding, measures of distortion based on image statistics may be disappointing, if properties of human visual system are not considered. Common problems are smoothed edges and different reconstruction qualities of different regions in the image. This study aims to achieve image coding with perceptually good reconstruction quality, by seeking solutions to the above problems. Proposed scheme, distortion equalizing coder (DEC), is based on contour- texture approach. Image is segmented, and, contour and...
Texture and edge preserving multiframe super-resolution Turgay, Emre; Akar, Gözde (Institution of Engineering and Technology (IET), 2014-09-01) Super-resolution (SR) image reconstruction refers to methods where a higher resolution image is reconstructed using a set of overlapping aliased low-resolution observations of the same scene. Although edge preservation has been a widely explored topic in SR literature, texture-specific regularisation has recently gained interest. In this study, texture-specific regularisation is handled as a post-processing step. A two stage method is proposed, comprising multiple SR reconstructions with different regularis...
Cognitive aspects of image zooming Öncül, Gaye; Tarı, Sibel; Tekman, Hasan Gürkan; Department of Cognitive Sciences (2002) In this work, digital image zooming methods and image distortion metrics are examined in terms of a quality criterion based on image appearance. Neurophysiological and psychophysical foundations of perception driven image distortion metrics are presented. Results of a comparative study based on both perceptual distortion metrics and conventional distortion metrics for standard zooming methods are discussed. Open problems in image quality definition as well as image enlargement methods are presented and impr...
Shape descriptors based on intersection consistency and global binary patterns Sivri, Erdal; Kalkan, Sinan; Department of Computer Engineering (2012) Shape description is an important problem in computer vision because most vision tasks that require comparing or matching visual entities rely on shape descriptors. In this thesis, two novel shape descriptors are proposed, namely Intersection Consistency Histogram (ICH) and Global Binary Patterns (GBP). The former is based on a local regularity measure called Intersection Consistency (IC), which determines whether edge pixels in an image patch point towards the center or not. The second method, called Globa...

Citation Formats

F. YAVUZ and S. Kalkan, “Segment Augmentation and Differentiable Ranking for Logo Retrieval,” Montreal, Kanada, 2022, vol. 2022-August, Accessed: 00, 2023. [Online]. Available: https://hdl.handle.net/11511/101848.