Andres Romero - Academia.edu (original) (raw)

Papers by Andres Romero

2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

Figure 1. We use Denoising Diffusion Probabilistic Models (DDPM) for inpainting. The process is c... more Figure 1. We use Denoising Diffusion Probabilistic Models (DDPM) for inpainting. The process is conditioned on the masked input (left). It starts from a random Gaussian noise sample that is iteratively denoised until it produces a high-quality output. Since this process is stochastic, we can sample multiple diverse outputs. The DDPM prior forces a harmonized image, is able to reproduce texture from other regions, and inpaint semantically meaningful content.

Proceedings of the Fifth International Conference on Informatics in Control, Automation and Robotics Service, 2008

In this paper we propose a very simple procedure for the design and analysis of low-pass and high... more In this paper we propose a very simple procedure for the design and analysis of low-pass and high-pass wave digital filters derived from reference filter given in the lattice configuration. Wave Digital Filters derived from reference filter in lattice configuration can be designed with excellent pass band properties. They can be proposed and implemented without the knowledge of classical filter theory. In this paper we present tables for Butterworth, Chebychev and Elliptic low-pass filter design. In the examples we demonstrate programs in MATLAB that permits analyze the attenuation properties of the designed filters. In the end of our article we realize wave digital filter using Embedded Target for Texas instruments TMS320C6000 DSP Platform. The model of the WDF was created by means of serial and parallel blocks that were added to the window Simulink Library Browser between common Used Blocks.

2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), 2021

Real-world Super-Resolution (SR) has been traditionally tackled by first learning a specific degr... more Real-world Super-Resolution (SR) has been traditionally tackled by first learning a specific degradation model that resembles the noise and corruption artifacts in lowresolution imagery. Thus, current methods lack generalization and lose their accuracy when tested on unseen types of corruption. In contrast to the traditional proposal, we present Robust Super-Resolution (RSR), a method that leverages the generalization capability of adversarial attacks to tackle real-world SR. Our novel framework poses a paradigm shift in the development of real-world SR methods. Instead of learning a dataset-specific degradation, we employ adversarial attacks to create difficult examples that target the model's weaknesses. Afterward, we use these adversarial examples during training to improve our model's capacity to process noisy inputs. We perform extensive experimentation on synthetic and real-world images and empirically demonstrate that our RSR method generalizes well across datasets without retraining for specific noise priors. By using a single robust model, we outperform state-of-theart specialized methods on real-world benchmarks.

2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021

Humans can communicate emotions through a plethora of facial expressions, each with its own inten... more Humans can communicate emotions through a plethora of facial expressions, each with its own intensity, nuances and ambiguities. The generation of such variety by means of conditional GANs is limited to the expressions encoded in the used label system. These limitations are caused either due to burdensome labelling demand or the confounded label space. On the other hand, learning from inexpensive and intuitive basic categorical emotion labels leads to limited emotion variability. In this paper, we propose a novel GAN-based framework that learns an expressive and interpretable conditional space (usable as a label space) of emotions, instead of conditioning on handcrafted labels. Our framework only uses the categorical labels of basic emotions to learn jointly the conditional space as well as emotion manipulation. Such learning can benefit from the image variability within discrete labels, especially when the intrinsic labels reside beyond the discrete space of the defined. Our experiments demonstrate the effectiveness of the proposed framework, by allowing us to control and generate a gamut of complex and compound emotions while using only the basic categorical emotion labels during training.

Advances in Physiology Education, 2021

Undergraduates use a spike sorting routine developed in Octave to analyze the spiking activity ge... more Undergraduates use a spike sorting routine developed in Octave to analyze the spiking activity generated from mechanical stimulation of spines of cockroach legs with the inexpensive SpikerBox amplifier and the free software Audacity. Students learn the procedures involved in handling the cockroaches and recording extracellular action potentials (spikes) with the SpikerBox apparatus as well as the importance of spike sorting for analysis in neuroscience. The spike sorting process requires students to choose the spike threshold and spike selection criteria and interact with the clustering process that forms the groups of similar spikes. Once the spike groups are identified, interspike intervals and neuron firing frequencies can be calculated and analyzed. A classic neurophysiology lab exercise is thus adapted to be interdisciplinary for underrepresented students in a small rural college.

2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2021

Shadow removal is an important computer vision task aiming at the detection and successful remova... more Shadow removal is an important computer vision task aiming at the detection and successful removal of the shadow produced by an occluded light source and a photorealistic restoration of the image contents. Decades of research produced a multitude of hand-crafted restoration techniques and, more recently, learned solutions from shadowed and shadow-free training image pairs. In this work, we propose a single image shadow removal solution via self-supervised learning by using a conditioned mask. We rely on self-supervision and jointly learn deep models to remove and add shadows to images. We derive two variants for learning from paired images and unpaired images, respectively. Our validation on the recently introduced ISTD and USR datasets demonstrate large quantitative and qualitative improvements over the state-of-the-art for both paired and unpaired learning settings.

Banca Espanola Revista Del Mundo Bancario, 1995

Teoria Y Practica De Las Publicaciones Infantiles Y Juveniles 1978 Isbn 84 7483 010 9 Pags 321 322, 1978

Teoria Y Practica De Las Publicaciones Infantiles Y Juveniles 1978 Isbn 84 7483 010 9 Pags 323 327, 1978

Se presenta un análisis del sistema de planeamiento de la producción industrial, en el caso de pl... more Se presenta un análisis del sistema de planeamiento de la producción industrial, en el caso de plantas distribuidas con productos complejos. En general el problema de la distribución física de la planta agrega varios elementos a la complejidad intrínseca del planeamiento de la producción, fundamentalmente por el tema de los recursos físicamente distantes (máquinas, materias primas, operarios), los tiempos de movimiento de los diferentes stocks así como la centralización/distribución del control de calidad. Por otra parte se aborda el caso de los "productos complejos", entendiendo por tales aquéllos en los que el Análisis de producción contiene no sólo materias primas, sino también productos semielaborados (en 1 o más niveles) cuya producción debe planificarse coherentemente con los productos "finales". Esta estructura requiere una descomposición en niveles de los recursos de stock/semielaborados requeridos para cumplir los planes de producción, con relaciones de interdependencia que requieren validar el orden de producción. Por último se discuten diferentes criterios de optimización de los planes de producción, analizando los resultados obtenidos con un enfoque basado en la reducción de complejidad en la gestión de planta.

This thesis proposes a computer vision system for detecting and tracking multiple targets in vide... more This thesis proposes a computer vision system for detecting and tracking multiple targets in videos. The covariance matching method is the guiding thread of our work because it offers a compact representation of the target by embedding heterogeneous features in a elegant way. Therefore, it is efficient both for tracking and recognition. Four categories of contributions are proposed. The first one deals with the adaptation to a changing context, following two aspects. A preliminary work consists in the adaptation of color according to lighting variations and relevance of the color. Then, literature shows a wide variety of tracking methods, which have both advantages and limitations, depending on the object to track and the context. Here, a deterministic method is developed to automatically adapt the tracking method to the context through the cooperation of two complementary techniques. A first proposition combines covariance matching for modeling characteristics texture-color informa...

Lecture Notes in Computer Science, 2013

This paper addresses the multi-target tracking problem with the help of a matching method where m... more This paper addresses the multi-target tracking problem with the help of a matching method where moving objects are detected in each frame, tracked when it is possible and matched by similarity of covariance matrices when difficulties arrive. Three contributions are proposed. First, a compact vector based on color invariants and Local Binary Patterns Variance is compared to more classical features vectors. To accelerate object re-identification, our second proposal is the use of a more efficient arrangement of the covariance matrices. Finally, a multiple-target algorithm with special attention in occlusion handling, merging and separation of the targets is analyzed. Our experiments show the relevance of the method, illustrating the trade-off that has to be made between distinctiveness, invariance and compactness of the features.

2009 IEEE International Conference on Ultra-Wideband, 2009

Page 1. Iterative coexistence approaches for non-coherent multi-band impulse radio UWB Hanns-Ulri... more Page 1. Iterative coexistence approaches for non-coherent multi-band impulse radio UWB Hanns-Ulrich Dehner ∗ , Andrés Romero ∗ , Holger Jäkel ∗ , Dennis Burgkhardt ∗ , Rainer Moorfeld , Friedrich K. Jondral ∗ , and Adolf Finger ...

In this paper we propose a multi-target tracking based on the tracking-by-detection paradigm. The... more In this paper we propose a multi-target tracking based on the tracking-by-detection paradigm. The problem is casted as a discrete association problem where a cost is assigned to each detection-tracklet pair and the evolution of many factors such as position, speed and appearance is observed. As new tracklets enter the scene their appearance is modeled using covariance matrices equipped with the total Bregman divergence to perform the comparisons and robust model updates. Our method provides near-state of the art results in terms of accuracy and is able to execute in real-time. Index Terms-Total Bregman divergence, tracking in Riemannian manifolds, multiple object tracking.

2009 IEEE Symposium on Computational Intelligence in Milti-Criteria Decision-Making, 2009

This paper examines two strategies in order to improve the performance of multi-objective evoluti... more This paper examines two strategies in order to improve the performance of multi-objective evolutionary algorithms when applied to problems with many objectives: informed initialization and extreme solutions sub-population. The informed initialization is the inclusion of approximations of extreme and internal points of the Pareto front in the initial population. These approximations, called informed initial solutions, are found using a high quality evolutionary or local search algorithm on single objective problems obtained by scalarizing the multiple goals into a single goal by the use of weight vectors. The extreme solutions sub-population is proposed here to keep the best approximations of the extreme points of the Pareto front at any point of the evolution, and the selection scheme is biased to give these solutions slightly higher chances of being selected. Experimental results applying these two strategies in continuous and combinatorial benchmark problems show that the diversity in the final solutions is improved, while preserving the proximity to the Pareto front. Some additional experiments that demonstrate how the number of initial informed solutions affects the performance are also presented.

2011 IEEE International Conference on Signal and Image Processing Applications (ICSIPA), 2011

This paper proposes a color tracking strategy designed to improve the robustness against luminanc... more This paper proposes a color tracking strategy designed to improve the robustness against luminance and saturation changes due to illumination variations. On the one hand, color is helpful in terms of photometric invariance and separability power. On the other hand, it is more costly in time and resources and most color invariants are ill-defined at low saturation. To answer these issues, the proposed method weights the color and luminance information adaptively with respect to the current color saturation. A particular attention is also given to chose a color representation with low computational cost. The method is compared to classical color tracking in terms of accuracy, robustness and executing times. The experimental results achieved on several image sequences confirm the good performances of the method.

2008 IEEE Congress on Evolutionary Computation (IEEE World Congress on Computational Intelligence), 2008

I. INTRODUCTION Recently, the generation of textual information has grown considerably; thus, cur... more I. INTRODUCTION Recently, the generation of textual information has grown considerably; thus, currently, many people and organizations need to work with huge amounts of data. Such data can be in the form of working papers, corporate documents and e-mail, among ...

Journal of Real-Time Image Processing, 2013

The paper proposes contributions for Mean-Shift (M S) and Covariance Tracking (CT), and makes the... more The paper proposes contributions for Mean-Shift (M S) and Covariance Tracking (CT), and makes these two complementary methods cooperate. While M S runs fast and can handle non-rigid objects represented by their color distribution, CT is more time-consuming but achieves a generic tracking by mixing color and texture information. Each method is modified in order to alleviate their intrinsic limitations, and make the tracking adaptive to a changing context. Concerning M S, the colorspace is changed automatically when necessary to enhance the distinction between the object and the background. Regarding CT , the number of features is reduced without loss of accuracy, by using Local Binary Patterns. Finally, their complementary advantages are exploited in a cooperation process, which runs faster than CT alone, and is more robust than M S alone. A comprehensive study is made for their acceleration and their efficient execution on different multi-core CPUs. A speedup of ×2.8 is reached for M S and ×2.6 for CT .