arXiv subjects

Facundo Quiroga

Publications and source records attributed to Facundo Quiroga.

5 recordsLinked to original sources

arXiv2026

Target-Side Paraphrase Augmentation for Sign Language Translation with Large Language Models

Sign language translation (SLT) remains constrained by the limited availability of paired sign-video/text corpora and by the heavy-tailed vocabularies typical of real-world datasets. We study a target-side augmentation strategy in which a large language model (LLM) generates controlled paraphrase variants of the reference spoken-language sentence while the sign input remains unchanged. Concretely, we use GPT-4o to produce semantically faithful variants of the training targets and train a Signformer-style pose-based Transformer under a two-stage schedule: pre-training on the augmented corpus followed by fine-tuning on the original references. We evaluate this strategy on three datasets that span complementary challenges: PHOENIX14T (German Sign Language), a real-world corpus with moderate lexical diversity; the Greek Sign Language Dataset with highly controlled, repetitive recordings; and LSA-T (Argentinian Sign Language), a naturalistic corpus with a large vocabulary and severe long-tail sparsity. This range allows us to characterize precisely when and why target-side augmentation is beneficial. On PHOENIX14T, augmentation improves BLEU-4 from 9.56 to 10.33, demonstrating that paraphrastic exposure helps the decoder generalize beyond memorized reference phrasing. The near-saturated GSL baseline and the extremely sparse LSA-T setting reveal the limits of the approach: in both cases, single-reference lexical overlap metrics are insufficient to capture the full picture, motivating a complementary semantic evaluation. To our knowledge, this is the first study to examine LLM-generated target-side paraphrases as an augmentation mechanism for SLT, and the first to apply an LLM-as-a-Judge evaluation protocol to SLT. This complementary evaluation reveals gains in semantic fidelity that lexical overlap metrics understate.

cs.CL↗

arXiv2025

HandCraft: Dynamic Sign Generation for Synthetic Data Augmentation

Sign Language Recognition (SLR) models face significant performance limitations due to insufficient training data availability. In this article, we address the challenge of limited data in SLR by introducing a novel and lightweight sign generation model based on CMLPe. This model, coupled with a synthetic data pretraining approach, consistently improves recognition accuracy, establishing new state-of-the-art results for the LSFB and DiSPLaY datasets using our Mamba-SL and Transformer-SL classifiers. Our findings reveal that synthetic data pretraining outperforms traditional augmentation methods in some cases and yields complementary benefits when implemented alongside them. Our approach democratizes sign generation and synthetic data pretraining for SLR by providing computationally efficient methods that achieve significant performance improvements across diverse datasets.

cs.CV↗

arXiv2025

Bringing Balance to Hand Shape Classification: Mitigating Data Imbalance Through Generative Models

Most sign language handshape datasets are severely limited and unbalanced, posing significant challenges to effective model training. In this paper, we explore the effectiveness of augmenting the training data of a handshape classifier by generating synthetic data. We use an EfficientNet classifier trained on the RWTH German sign language handshape dataset, which is small and heavily unbalanced, applying different strategies to combine generated and real images. We compare two Generative Adversarial Networks (GAN) architectures for data generation: ReACGAN, which uses label information to condition the data generation process through an auxiliary classifier, and SPADE, which utilizes spatially-adaptive normalization to condition the generation on pose information. ReACGAN allows for the generation of realistic images that align with specific handshape labels, while SPADE focuses on generating images with accurate spatial handshape configurations. Our proposed techniques improve the current state-of-the-art accuracy on the RWTH dataset by 5%, addressing the limitations of small and unbalanced datasets. Additionally, our method demonstrates the capability to generalize across different sign language datasets by leveraging pose-based generation trained on the extensive HaGRID dataset. We achieve comparable performance to single-source trained classifiers without the need for retraining the generator.

cs.CV↗

arXiv2022

LSA-T: The first continuous Argentinian Sign Language dataset for Sign Language Translation

Sign language translation (SLT) is an active field of study that encompasses human-computer interaction, computer vision, natural language processing and machine learning. Progress on this field could lead to higher levels of integration of deaf people. This paper presents, to the best of our knowledge, the first continuous Argentinian Sign Language (LSA) dataset. It contains 14,880 sentence level videos of LSA extracted from the CN Sordos YouTube channel with labels and keypoints annotations for each signer. We also present a method for inferring the active signer, a detailed analysis of the characteristics of the dataset, a visualization tool to explore the dataset and a neural SLT model to serve as baseline for future experiments.

cs.CV↗

arXiv2020

Reddening-free Q indices to identify Be star candidates

Astronomical databases currently provide high-volume spectroscopic and photometric data. While spectroscopic data is better suited to the analysis of many astronomical objects, photometric data is relatively easier to obtain due to shorter telescope usage time. Therefore, there is a growing need to use photometric information to automatically identify objects for further detailed studies, specially Hα emission line stars such as Be stars. Photometric color-color diagrams (CCDs) are commonly used to identify this kind of objects. However, their identification in CCDs is further complicated by the reddening effect caused by both the circumstellar and interstellar gas. This effect prevents the generalization of candidate identification systems. Therefore, in this work we evaluate the use of neural networks to identify Be star candidates from a set of OB-type stars. The networks are trained using a labeled subset of the VPHAS+ and 2MASS databases, with filters u, g, r, Hα, i, J, H, and K. In order to avoid the reddening effect, we propose and evaluate the use of reddening-free Q indices to enhance the generalization of the model to other databases and objects. To test the validity of the approach, we manually labeled a subset of the database, and use it to evaluate candidate identification models. We also labeled an independent dataset for cross dataset evaluation. We evaluate the recall of the models at a 99% precision level on both test sets. Our results show that the proposed features provide a significant improvement over the original filter magnitudes.

astro-ph.SR↗