Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 9
Encontrada(s) 1 página(s)

Repositorio Digital de la UPF
Publicación Artículo científico (JournalArticle)

Multilabel prototype generation for data reduction in k-nearest neighbour classification

  • Valero-Mas, Jose J.
  • Gallego, Antonio Javier
  • Alonso-Jiménez, Pablo
  • Serra, Xavier
Prototype Generation (PG) methods are typically considered for improving the efficiency of the k-Nearest Neighbour (kNN) classifier when tackling high-size corpora. Such approaches aim at generating a reduced version of the corpus without decreasing the classification performance when compared to the initial set. Despite their large application in multiclass scenarios, very few works have addressed the proposal of PG methods for the multilabel space. In this regard, this work presents the novel adaptation of four multiclass PG strategies to the multilabel case. These proposals are evaluated with three multilabel kNN-based classifiers, 12 corpora comprising a varied range of domains and corpus sizes, and different noise scenarios artificially induced in the data. The results obtained show that the proposed adaptations are capable of significantly improving—both in terms of efficiency and classification performance—the only reference multilabel PG work in the literature as well as the case in which no PG method is applied, also presenting statistically superior robustness in noisy scenarios. Moreover, these novel PG strategies allow prioritising either the efficiency or efficacy criteria through its configuration depending on the target scenario, hence covering a wide area in the solution space not previously filled by other works., This research was partially funded by the Spanish Ministerio de Ciencia e Innovación through the MultiScore (PID2020-118447RA-I00) and DOREMI (TED2021-132103A-I00) projects. The first author is supported by grant APOSTD/2020/256 from “Programa I+D+i de la Generalitat Valenciana”.
Repositorio Digital de la UPF
Publicación Contribución de congreso (ConferenceOutput)

Addressing class imbalance in multilabel Prototype Generation for k-Nearest Neighbor classification

  • Penarrubia, Carlos
  • Valero-Mas, Jose J.
  • Gallego, Antonio Javier
  • Calvo-Zaragoza, Jorge
Comunicació presentada a 11th Iberian Conference (IbPRIA 2023), celebrada del 27 al 30 de juny de 2023 a Alacant, Espanya., Prototype Generation (PG) methods seek to improve the efficiency of the k-Nearest Neighbor (kNN) classifier by obtaining a reduced version of a given reference dataset following certain heuristics. Despite being largely addressed topic in multiclass scenarios, few works deal with PG in multilabel environments. Hence, the existing proposals exhibit a number of limitations, being label imbalance one of paramount relevance as it constitutes a typical challenge of multilabel datasets. This work proposes two novel merging policies for multilabel PG schemes specifically devised for label imbalance, as well as a mechanism to prevent inappropriate samples from undergoing a reduction process. These proposals are applied to three existing multilabel PG methods—Multilabel Reduction through Homogeneous Clustering, Multilabel Chen, and Multilabel Reduction through Space Partitioning—and evaluated on 12 different data assortments with different degrees of label imbalance. The results prove that the proposals overcome—in some cases in a significant manner—those obtained with the original methods, hence validating the presented approaches and enabling further research lines on this topic., This work was supported by the I+D+i project TED2021-132103A-I00 (DOREMI), funded by MCIN/AEI/10.13039/501100011033.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

An overview of ensemble and feature learning in few-shot image classification using siamese networks

  • Valero-Mas, Jose J.
  • Gallego, Antonio-Javier
  • Rico-Juan, Juan Ramón
  • Gallego, Antonio-Javier|||/items/8f865afc-8f63-4e0d-9408-bebc7ba836d0
  • Rico-Juan, Juan Ramón|||/items/ab906fc8-ea97-4172-a38d-6eb429e9dc84
  • Valero-Mas, Jose J.|||/items/8ad448df-2d0f-41e8-8403-cd85202b7263
Siamese Neural Networks (SNNs) constitute one of the most representative approaches for addressing Few-Shot Image Classification. These schemes comprise a set of Convolutional Neural Network (CNN) models whose weights are shared across the network, which results in fewer parameters to train and less tendency to overfit. This fact eventually leads to better convergence capabilities than standard neural models when considering scarce amounts of data. Based on a contrastive principle, the SNN scheme jointly trains these inner CNN models to map the input image data to an embedded representation that may be later exploited for the recognition process. However, in spite of their extensive use in the related literature, the representation capabilities of SNN schemes have neither been thoroughly assessed nor combined with other strategies for boosting their classification performance. Within this context, this work experimentally studies the capabilities of SNN architectures for obtaining a suitable embedded representation in scenarios with a severe data scarcity, assesses the use of train data augmentation for improving the feature learning process, introduces the use of transfer learning techniques for further exploiting the embedded representations obtained by the model, and uses test data augmentation for boosting the performance capabilities of the SNN scheme by mimicking an ensemble learning process. The results obtained with different image corpora report that the combination of the commented techniques achieves classification rates ranging from 69% to 78% with just 5 to 20 prototypes per class whereas the CNN baseline considered is unable to converge. Furthermore, upon the convergence of the baseline model with the sufficient amount of data, still the adequate use of the studied techniques improves the accuracy in figures from 4% to 9%.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Multilabel Prototype Generation for data reduction in K-Nearest Neighbour classification

  • Valero-Mas, Jose J.
  • Gallego, Antonio-Javier
  • Alonso-Jiménez, Pablo
  • Serra, Xavier
  • Gallego, Antonio-Javier|||/items/8f865afc-8f63-4e0d-9408-bebc7ba836d0
  • Valero-Mas, Jose J.|||/items/8ad448df-2d0f-41e8-8403-cd85202b7263
Prototype Generation (PG) methods are typically considered for improving the efficiency of the k-Nearest Neighbour (kNN) classifier when tackling high-size corpora. Such approaches aim at generating a reduced version of the corpus without decreasing the classification performance when compared to the initial set. Despite their large application in multiclass scenarios, very few works have addressed the proposal of PG methods for the multilabel space. In this regard, this work presents the novel adaptation of four multiclass PG strategies to the multilabel case. These proposals are evaluated with three multilabel kNN-based classifiers, 12 corpora comprising a varied range of domains and corpus sizes, and different noise scenarios artificially induced in the data. The results obtained show that the proposed adaptations are capable of significantly improving—both in terms of efficiency and classification performance—the only reference multilabel PG work in the literature as well as the case in which no PG method is applied, also presenting statistically superior robustness in noisy scenarios. Moreover, these novel PG strategies allow prioritising either the efficiency or efficacy criteria through its configuration depending on the target scenario, hence covering a wide area in the solution space not previously filled by other works.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

A Preliminary Study of Few-shot Learning for Layout Analysis of Music Scores

  • Castellanos, Francisco J.
  • Gallego, Antonio-Javier
  • Fujinaga, Ichiro
  • Castellanos, Francisco J.|||/items/6b515109-60a8-4a00-a12d-0726355ae966
  • Gallego, Antonio-Javier|||/items/8f865afc-8f63-4e0d-9408-bebc7ba836d0
Few-shot techniques offer a promising avenue to reduce the high demand for annotated data required by current machine learning-based applications, such as Optical Music Recognition (OMR). This is a field dedicated to the automatic transcription of music notation from sheet music images. Traditional OMR systems strongly depend on layout analysis, a crucial step involving the identification and segmentation of several components within a music score, such as staff lines, text, or notes. The standard approach requires extensive fully annotated training data, which are resource-intensive and time-consuming to label and curate by domain experts. We present a preliminary study on the use of few-shot learning to alleviate the disadvantages associated with manual annotations. The proposal minimizes the human effort required by employing only partial annotations. For this, we introduce an oversampling technique to train models using a limited set of annotated patches extracted from the score images. Our experimental findings, conducted on four benchmark datasets, underscore the efficacy of the proposed patch extraction. Despite operating with a reduced amount of annotated data, our method achieves performance levels competitive with models trained on the complete dataset. This work points out the potential of few-shot learning in the context of layout analysis for music scores, offering the promise of more efficient and accessible OMR systems.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Efficient multi-task progressive learning for semantic segmentation and disparity estimation

  • Cuevas-Velasquez, Hanz
  • Galán Cuenca, Alejandro
  • Fisher, Robert B.
  • Gallego, Antonio-Javier
  • Gallego, Antonio-Javier|||/items/8f865afc-8f63-4e0d-9408-bebc7ba836d0
  • Galán Cuenca, Alejandro|||/items/d81019ac-5845-4e4f-8af5-9c00dc6d1467
Scene understanding is an important area in robotics and autonomous driving. To accomplish these tasks, the 3D structures in the scene have to be inferred to know what the objects and their locations are. To this end, semantic segmentation and disparity estimation networks are typically used, but running them individually is inefficient since they require high-performance resources. A possible solution is to learn both tasks together using a multi-task approach. Some current methods address this problem by learning semantic segmentation and monocular depth together. However, monocular depth estimation from single images is an ill-posed problem. A better solution is to estimate the disparity between two stereo images and take advantage of this additional information to improve the segmentation. This work proposes an efficient multi-task method that jointly learns disparity and semantic segmentation. Employing a Siamese backbone architecture for multi-scale feature extraction, the method integrates specialized branches for disparity estimation and coarse and refined segmentations, leveraging progressive task-specific feature sharing and attention mechanisms to enhance accuracy for solving both tasks concurrently. The proposal achieves state-of-the-art results for joint segmentation and disparity estimation on three distinct datasets: Cityscapes, TrimBot2020 Garden, and S-ROSeS, using only of the parameters of previous approaches.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Kurcuma: a kitchen utensil recognition collection for unsupervised domain adaptation

  • Rosello, Adrian
  • Valero-Mas, Jose J.
  • Gallego, Antonio-Javier
  • Saez-Perez, Javier
  • Calvo-Zaragoza, Jorge
  • Calvo-Zaragoza, Jorge|||/items/a4f3686d-f0fe-4317-a73c-cd711ceb1dde
  • Gallego, Antonio-Javier|||/items/8f865afc-8f63-4e0d-9408-bebc7ba836d0
  • Rosello, Adrian|||/items/b80e2844-9518-476b-821c-83d99598ffb3
  • Valero-Mas, Jose J.|||/items/8ad448df-2d0f-41e8-8403-cd85202b7263
The use of deep learning makes it possible to achieve extraordinary results in all kinds of tasks related to computer vision. However, this performance is strongly related to the availability of training data and its relationship with the distribution in the eventual application scenario. This question is of vital importance in areas such as robotics, where the targeted environment data are barely available in advance. In this context, domain adaptation (DA) techniques are especially important to building models that deal with new data for which the corresponding label is not available. To promote further research in DA techniques applied to robotics, this work presents Kurcuma (Kitchen Utensil Recognition Collection for Unsupervised doMain Adaptation), an assortment of seven datasets for the classification of kitchen utensils—a task of relevance in home-assistance robotics and a suitable showcase for DA. Along with the data, we provide a broad description of the main characteristics of the dataset, as well as a baseline using the well-known domain-adversarial training of neural networks approach. The results show the challenge posed by DA on these types of tasks, pointing to the need for new approaches in future work.
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Multilabel prototype generation for data reduction in k-nearest neighbour classification

  • Valero-Mas, Jose J.
  • Gallego, Antonio Javier
  • Alonso-Jiménez, Pablo
  • Serra, Xavier
Prototype Generation (PG) methods are typically considered for improving the efficiency of the k-Nearest Neighbour (kNN) classifier when tackling high-size corpora. Such approaches aim at generating a reduced version of the corpus without decreasing the classification performance when compared to the initial set. Despite their large application in multiclass scenarios, very few works have addressed the proposal of PG methods for the multilabel space. In this regard, this work presents the novel adaptation of four multiclass PG strategies to the multilabel case. These proposals are evaluated with three multilabel kNN-based classifiers, 12 corpora comprising a varied range of domains and corpus sizes, and different noise scenarios artificially induced in the data. The results obtained show that the proposed adaptations are capable of significantly improving—both in terms of efficiency and classification performance—the only reference multilabel PG work in the literature as well as the case in which no PG method is applied, also presenting statistically superior robustness in noisy scenarios. Moreover, these novel PG strategies allow prioritising either the efficiency or efficacy criteria through its configuration depending on the target scenario, hence covering a wide area in the solution space not previously filled by other works., This research was partially funded by the Spanish Ministerio de Ciencia e Innovación through the MultiScore (PID2020-118447RA-I00) and DOREMI (TED2021-132103A-I00) projects. The first author is supported by grant APOSTD/2020/256 from “Programa I+D+i de la Generalitat Valenciana”.
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Addressing class imbalance in multilabel Prototype Generation for k-Nearest Neighbor classification

  • Penarrubia, Carlos
  • Valero-Mas, Jose J.
  • Gallego, Antonio Javier
  • Calvo-Zaragoza, Jorge
Comunicació presentada a 11th Iberian Conference (IbPRIA 2023), celebrada del 27 al 30 de juny de 2023 a Alacant, Espanya., Prototype Generation (PG) methods seek to improve the efficiency of the k-Nearest Neighbor (kNN) classifier by obtaining a reduced version of a given reference dataset following certain heuristics. Despite being largely addressed topic in multiclass scenarios, few works deal with PG in multilabel environments. Hence, the existing proposals exhibit a number of limitations, being label imbalance one of paramount relevance as it constitutes a typical challenge of multilabel datasets. This work proposes two novel merging policies for multilabel PG schemes specifically devised for label imbalance, as well as a mechanism to prevent inappropriate samples from undergoing a reduction process. These proposals are applied to three existing multilabel PG methods—Multilabel Reduction through Homogeneous Clustering, Multilabel Chen, and Multilabel Reduction through Space Partitioning—and evaluated on 12 different data assortments with different degrees of label imbalance. The results prove that the proposals overcome—in some cases in a significant manner—those obtained with the original methods, hence validating the presented approaches and enabling further research lines on this topic., This work was supported by the I+D+i project TED2021-132103A-I00 (DOREMI), funded by MCIN/AEI/10.13039/501100011033.