Search results for: triplet%20network

Authors: Wentian Shi, Daming Shi, Maysam Orouskhani, Feng Tian

Abstract:

Currently the most prevalent deep learning methods require a large amount of data for training, whereas few-shot learning tries to learn a model from limited data without extensive retraining. In this paper, we present a loss function based on triplet loss for solving few-shot problem using metric based learning. Instead of setting the margin distance in triplet loss as a constant number empirically, we propose an adaptive margin distance strategy to obtain the appropriate margin distance automatically. We implement the strategy in the deep siamese network for deep metric embedding, by utilizing an optimization approach by penalizing the worst case and rewarding the best. Our experiments on image recognition and co-segmentation model demonstrate that using our proposed triplet loss with adaptive margin distance can significantly improve the performance.

Keywords: Few-shot learning, triplet network, adaptive margin, deep learning.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 818

Authors: Berna Istegun, Erkan Celebi

Abstract:

The main objective of this experimental study is to assess the shear strength and the crack behavior of the triplets built of perforated brickwork masonry elements. In order to observe the influence of shear resistance and energy dissipating before and after retrofitting applications by using the reinforcing system, static-cyclic shear tests were employed in the structural mechanics laboratory of Sakarya University. The reinforcing system is composed of hybrid multiaxial seismic fabric consisting of alkali resistant glass and polypropylene fibers. The plaster as bonding material used in the specimen’s retrofitting consists of expanded glass granular. In order to acquire exact measuring data about the failure behavior of the two mortar joints under shear stressing, vertical load-controlled cylinder having force capacity of 50 kN and loading rate of 1.5 mm/min. with an internal inductive displacement transducers is carried out perpendicular to the triplet specimens. In this study, a total of six triplet specimens with textile reinforcement were prepared for these shear bond tests. The three of them were produced as single-sided reinforced triplets with seismic fabric, while the others were strengthened on both sides. In addition, three triplet specimens without retrofitting and plaster were also tested as reference samples. The obtained test results were given in the manner of force-displacement relationships, ductility coefficients and shear strength parameters comparatively. It is concluded that two-side seismic textile applications on masonry elements with relevant plaster have considerably increased the sheer force resistance and the ductility capacity.

Keywords: Triplet shears tests, retrofitting, seismic fabric, perforated brickwork, expanded glass granular.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1238

Authors: T. Song, T. Shimamoto

Abstract:

In this paper a fast motion estimation method for H.264/AVC named Triplet Search Motion Estimation (TS-ME) is proposed. Similar to some of the traditional fast motion estimation methods and their improved proposals which restrict the search points only to some selected candidates to decrease the computation complexity, proposed algorithm separate the motion search process to several steps but with some new features. First, proposed algorithm try to search the real motion area using proposed triplet patterns instead of some selected search points to avoid dropping into the local minimum. Then, in the localized motion area a novel 3-step motion search algorithm is performed. Proposed search patterns are categorized into three rings on the basis of the distance from the search center. These three rings are adaptively selected by referencing the surrounding motion vectors to early terminate the motion search process. On the other hand, computation reduction for sub pixel motion search is also discussed considering the appearance probability of the sub pixel motion vector. From the simulation results, motion estimation speed improved by a factor of up to 38 when using proposed algorithm than that of the reference software of H.264/AVC with ignorable picture quality loss.

Keywords: Motion estimation, VLSI, image processing, search patterns

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1289

Authors: Muhammad Azam, Micheal Olaolu Arowolo, Fei He, Mihail Popescu, Dong Xu

Abstract:

The number of biological papers is growing quickly, which means that the number of biological pathway figures in those papers is also increasing quickly. Each pathway figure shows extensive biological information, like the names of genes and how the genes are related. However, manually annotating pathway figures takes a lot of time and work. Even though using advanced image understanding models could speed up the process of curation, these models still need to be made more accurate. To improve gene name recognition from pathway figures, we applied a Siamese network to map image segments to a library of pictures containing known genes in a similar way to person recognition from photos in many photo applications. We used a triple loss function and a triplet spatial pyramid pooling network by combining the triplet convolution neural network and the spatial pyramid pooling (TSPP-Net). We compared VGG19 and VGG16 as the Siamese network model. VGG16 achieved better performance with an accuracy of 93%, which is much higher than Optical Character Recognition (OCR) results.

Keywords: Biological pathway, image understanding, gene name recognition, object detection, Siamese network, Visual Geometry Group.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 550

Authors: Ankit Sinha, Soham Banerjee, Pratik Chattopadhyay

Abstract:

Automated product recognition in retail stores is an important real-world application in the domain of Computer Vision and Pattern Recognition. In this paper, we consider the problem of automatically identifying the classes of the products placed on racks in retail stores from an image of the rack and information about the query/product images. We improve upon the existing approaches in terms of effectiveness and memory requirement by developing a two-stage object detection and recognition pipeline comprising of a Faster-RCNN-based object localizer that detects the object regions in the rack image and a ResNet-18-based image encoder that classifies  the detected regions into the appropriate classes. Each of the models is fine-tuned using appropriate data sets for better prediction and data augmentation is performed on each query image to prepare an extensive gallery set for fine-tuning the ResNet-18-based product recognition model. This encoder is trained using a triplet loss function following the strategy of online-hard-negative-mining for improved prediction. The proposed models are lightweight and can be connected in an end-to-end manner during deployment to automatically identify each product object placed in a rack image. Extensive experiments using Grozi-32k and GP-180 data sets verify the effectiveness of the proposed model.

Keywords: Retail stores, Faster-RCNN, object localization, ResNet-18, triplet loss, data augmentation, product recognition.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 475

Authors: Robersy Sánchez, Ricardo Grau

Abstract:

A plausible architecture of an ancient genetic code is derived from an extended base triplet vector space over the Galois field of the extended base alphabet {D, G, A, U, C}, where the letter D represent one or more hypothetical bases with unspecific pairing. We hypothesized that the high degeneration of a primeval genetic code with five bases and the gradual origin and improvements of a primitive DNA repair system could make possible the transition from the ancient to the modern genetic code. Our results suggest that the Watson-Crick base pairing and the non-specific base pairing of the hypothetical ancestral base D used to define the sum and product operations are enough features to determine the coding constraints of the primeval and the modern genetic code, as well as the transition from the former to the later. Geometrical and algebraic properties of this vector space reveal that the present codon assignment of the standard genetic code could be induced from a primeval codon assignment. Besides, the Fourier spectrum of the extended DNA genome sequences derived from the multiple sequence alignment suggests that the called period-3 property of the present coding DNA sequences could also exist in the ancient coding DNA sequences.

Keywords: Genetic code vector space, primeval genetic code, power spectrum.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2329