María Elena Buemi

Juan Ignacio Bustos Gorostegui, Maria Elena Buemi

Egocentric action recognition is a challenging task due to erratic camera motion, frequent hand occlusion, and the difficulty of maintaining consistent visual representations over time. In this work, we propose a cross-modal architecture that combines RGB video and temporal hand skeleton data within a unified Mamba-based framework, exploiting the linear time complexity of State Space Models (SSMs). Our architecture consists of three components: a VideoMamba module for visual feature extraction, a skeleton encoder built on a stack of Mamba blocks, and a fusion module that integrates both modalities into a single representation. A central contribution of this work is the design and evaluation of four Class (CLS) token mixing strategies for multimodal fusion: Naive, Average, Weighted and Context-based. These strategies differ in how the pretrained unimodal CLS tokens, which role is to act as information sinks concentrating learned representations, are leveraged to initialize the mixed CLS token used for final classification. We evaluate all strategies on the H2O dataset. Experimental results show that the Average strategy achieves the best performance, yielding gains of over 10% Top-1 accuracy in the Tiny configuration and 2% in the Small configuration over the VideoMamba baseline.

Nicolas Mastropasqua, Ignacio Bugueno-Cordova, Rodrigo Verschae et al.

Micro-expression analysis has applications in domains such as Human-Robot Interaction and Driver Monitoring Systems. Accurately capturing subtle and fast facial movements remains difficult when relying solely on RGB cameras, due to limitations in temporal resolution and sensitivity to motion blur. Event cameras offer an alternative, with microsecond-level precision, high dynamic range, and low latency. However, public datasets featuring event-based recordings of Action Units are still scarce. In this work, we introduce a novel, preliminary multi-resolution and multi-modal micro-expression dataset recorded with synchronized RGB and event cameras under variable lighting conditions. Two baseline tasks are evaluated to explore the spatial-temporal dynamics of micro-expressions: Action Unit classification using Spiking Neural Networks (51.23\% accuracy with events vs. 23.12\% with RGB), and frame reconstruction using Conditional Variational Autoencoders, achieving SSIM = 0.8513 and PSNR = 26.89 dB with high-resolution event input. These promising results show that event-based data can be used for micro-expression recognition and frame reconstruction.

María Elena Buemi, Alejandro C. Frery, Heitor S. Ramos

Stack filters are a special case of non-linear filters. They have a good performance for filtering images with different types of noise while preserving edges and details. A stack filter decomposes an input image into stacks of binary images according to a set of thresholds. Each binary image is then filtered by a Boolean function, which characterizes the filter. Adaptive stack filters can be computed by training using a prototype (ideal) image and its corrupted version, leading to optimized filters with respect to a loss function. In this work we propose the use of training with selected samples for the estimation of the optimal Boolean function. We study the performance of adaptive stack filters when they are applied to speckled imagery, in particular to Synthetic Aperture Radar (SAR) images. This is done by evaluating the quality of the filtered images through the use of suitable image quality indexes and by measuring the classification accuracy of the resulting images. We used SAR images as input, since they are affected by speckle noise that makes classification a difficult task.

Maria E. Buemi, Marta Mejail, Julio Jacobo et al.

Stack filters are a special case of non-linear filters. They have a good performance for filtering images with different types of noise while preserving edges and details. A stack filter decomposes an input image into several binary images according to a set of thresholds. Each binary image is then filtered by a Boolean function, which characterizes the filter. Adaptive stack filters can be designed to be optimal; they are computed from a pair of images consisting of an ideal noiseless image and its noisy version. In this work we study the performance of adaptive stack filters when they are applied to Synthetic Aperture Radar (SAR) images. This is done by evaluating the quality of the filtered images through the use of suitable image quality indexes and by measuring the classification accuracy of the resulting images.