Template matching in the context of convolutional neural networks
| advisor: | RNDr. Jan Kalina, Ph.D. |
| e-mail: | show e-mail |
| type: | phd thesis |
| branch of study: | MINF, APIN |
| key words: | Templates, convolutional neural networks, deep learning, image analysis. |
| link: | http://www.cs.cas.cz/staff/kalina |
| description: | Methods based on templates (centroids) have been in spite of their simplicity widely used for object localization in images. Some recent works in image analysis have exploited template matching in a deep feature space produced by a convolutional neural network (CNN). The thesis will start with an overview of the literature on template-based methods in deep feature spaces. The main contribution is planned to improve the available naïve template-based approach for the context of deep learning. Particularly, the focus will be paid to the aspects of optimal construction, sparsity, robustness, or efficient computation of templates within deep learning. The student will implement the proposed approaches and apply them on suitable datasets of 2D images. |
| references: | [1] Gao B., Spratling M.W. (2021): Robust template matching via hierarchical convolutional features from a shape biased CNN. ArXiv:2007.15817. [2] Kalina J., Matonoha C. (2020): A sparse pair-preserving centroid-based supervised learning method for high-dimensional biomedical data or images. Biocybernetics and Biomedical Engineering 40(2), 774‒786. [3] Ferrari C., Berretti S., Bimbo A.D. (2019): Discovering identity specific activation patterns in deep descriptors for template based face recognition. 14th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2019), 1‒5. [4] Liu Y., Ling J., Liu Z., Shen J., Gao C. (2018): Finger vein secure biometric template generation based on deep learning. Soft Computing 22, 2257-2265 |
| last update: | 17.01.2023 14:07:02 |
administrator for this page:
Ľubomíra Dvořáková | last update: 09/12/2011