Browsing by Author "Delpiano, Jose"

Now showing 1 - 6 of 6
  • Thumbnail Image
    Item
    33
    (33, 33) Arevalo-Ramirez, Tito; Alfaro, Anali; Saavedra, Jose M.; Recabarren, Matias; Ponce-Donoso, Mauricio; Delpiano, Jose; 33
    Deep learning has gained popularity in recent years for reconstructing hyperspectral and multispectral images, offering cost-effective solutions and promising results. Research on hyperspectral image reconstruction feeds deep learning models with images at specific wavelengths and outputs images in other spectral bands. Although encouraging results of previous works, it should be determined to what extent the reconstructed information can lead to an advantage over the captured images. In this context, the present work inspects whether or not reconstructed spectral images add relevant information to segmentation networks for improving urban tree identification. Specifically, we generate red-edge (ReD) and near-infrared (NIR) images from RGB images using a conditional Generative Adversarial Network (cGAN). The training and validation are carried out with 5770 multispectral images obtained after a custom data augmentation process using an urban hyperspectral dataset. The testing outcomes reveal that ReD and NIR can be generated with an average Structural Similarity Index Measure of 0.93 and 0.88, respectively. Next, the cGAN generates ReD and NIR information of two RGB-based urban tree datasets (i.e., Jekyll, 3949 samples, and Arbocensus, 317 samples). Subsequently, DeepLabV3 and SegFormer segmentation networks are trained, validated, and tested using RGB, RGB+ReD, and RGB+NIR images from Jekyll and Arbocensus datasets. The experiments show that reconstructed multispectral images might not add information to segmentation networks that enhance their performance. Specifically, the p-values from a T-test show no significant difference between the performance of segmentation networks.
  • Thumbnail Image
    Item
    33
    (33, 33) Guamán-Rivera, Robert; Delpiano, Jose; Verschae, Rodrigo; 33
    Developing new convolutional neural network architectures and event-based camera representations could play a crucial role in autonomous navigation, pose estimation, and visual odometry applications. This study explores the potential of event cameras in optical flow estimation using convolutional neural networks. We provide a detailed description of the principles of operation and the software available for extracting and processing information from event cameras, along with the various event representation methods offered by this technology. Likewise, we identify four method categories to estimate optical flow using event cameras: gradient-based, frequency-based, correlation-based and neural network models. We report on these categories, including their latest developments, current status and challenges. We provide information on existing datasets and identify the appropriate dataset to evaluate deep learning-based optical flow estimation methods. We evaluate the accuracy of the implemented methods using the average endpoint error metric; meanwhile, the efficiency of the algorithms is evaluated as a function of execution time. Finally, we discuss research directions that promise future advances in this field.
  • Thumbnail Image
    Item
    33
    (33, 33) Moreno-Casas, Patricio A.; Scott, Felipe; Delpiano, Jose; Abell, Jose A.; Caicedo, Francisco; Muñoz, Raul; Vergara-Fernández, Alberto; 33
    The gas?liquid mass transfer coefficient is a key parameter to the design and operation of biotrickling filters that governs the transport rate of contaminants and oxygen from the gas phase to the liquid phase, where pollutant biodegradation occurs. Mass transfer coefficients are typically estimated via experimental procedures to produce empirical correlations, which are only valid for the bioreactor configuration and range of operational conditions under investigation. In this work, a new method for the estimation of the gas?liquid mass transfer coefficient in biotrickling filters is presented. This novel methodology couples a realistic description of the packing media (polyurethane foam without a biofilm) obtained using microtomography with computational fluid dynamics. The two-dimensional analysis reported in this study allowed capturing the mechanisms of the complex processes involved in the creeping porous air and water flow in the presence of capillary effects in biotrickling filters. Model predictions matched the experimental mass transfer coefficients (±30%) under a wide range of operational conditions.
  • Thumbnail Image
    Item
    33
    (33, 33) Aguayo, Paulo; Seriani, Sebastian; Delpiano, Jose; Farias, Gonzalo; Fujiyama, Taku; Velastin, Sergio A.; 33
    The platform–train interface (PTI) is considered a complex space where most interactions occur between passengers boarding and alighting. These interactions are critical under crowded conditions, affecting the experience of traveling and therefore the quality of life. The problem is that urban railway operators do not know what the density at the PTI is in real time, and therefore it is not possible to obtain a measure of the personal space of passengers boarding and alighting the train. To address this problem, a new method is developed to estimate the density of passengers on urban railway platforms using laboratory experiments. In those experiments, the use of computer vision is attractive, through the training of neural networks and image processing. The experiments considered a mock-up of a train carriage and its adjacent platform. In the boarding process, the results showed that the density using Voronoi polygons reached up to a 300% difference compared to the average values of density using Fruin’s Level of Service. However, in the case of alighting, that difference reached about 142% due to the space available for wheelchair users who needed assistance. These results would help practitioners to know where passengers are located at the PTI and, therefore, which part of the platform is more congested, requiring the implementation of crowd management measures in real time. Further studies need to include other types of passengers and different situations in existing stations.
  • Thumbnail Image
    Item
    33
    (SPIE, 33) Alzola, Rodrigo P.; Delpiano, Jose; Gajjar, Anuj; Stureborg, Rickard; Siadat, Seyed Mohammad; Ruberti, Jeffrey W.; Dimarzio, Charles A.; 33
    Collagen is one of the most important proteins in mammals, conforming most animal tissues. This work explores how a basic collagen monomer unit is visualized using fluorescence microscopy and how its spatial orientation is determined. Defining the orientation of collagen monomers is not a trivial problem, as the particle has a weak contrast and is relatively small. Possible attach fluorescence tags for contrast, but the size is still a problem for detecting orientation using fluorescence microscopy. This document presents a simulation of the visualization of collagen monomers and two methods for detecting monomer and classifying its orientation. A modify Gabor filter set, and an automatic classifier, trained by convolutional neuronal network (CNN), were used. By evaluating the performance of these two approaches compare to human observation, our results show that it is possible to determine the location and orientation of a single monomer with fluorescence microscopy. These findings can contribute to understanding collagen elements as collagen fibril.
  • Thumbnail Image
    Item
    Classification of multiple-state OAM superpositions using convolutional neural networks
    Oelckers, Tomas; Cisternas, Jaime E.; Delpiano, Jose