Romen Neris

Romen Neris

Auto-generated profile, based on publicly available information

Romen Neris

Institute for Applied Microelectronic (IUMA), University of Las Palmas de Gran Canaria, Gran Canaria, Spain

Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 April 2022 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Verified
Journal Article
Open access
RNRomen Neris
Romen Neris
ARAdrián Rodriguez-Molina
Adrián Rodriguez-Molina
RGRaul Guerra
Raul Guerra
Sebastian LopezSebastian Lopez
Sebastian Lopez
RSRoberto Sarmiento
Roberto Sarmiento
Over the last years, convolutional neural networks (CNNs) have been widely used in remote sensing applications, such as marine surveillance, traffic management, or road networks detection. However, since CNNs have extremely high computational, bandwith, and memory requirements, the hardware implementation of a CNN on space-grade devices like field-programmable gate array (FPGAs) for the on-board processing of the acquired images has brought many challenges, since the computational capabilities of the on-board hardware devices are limited. Hence, implementations have to be carefully planned. In this article, the authors present their work toward the implementation of an efficient CNN onto a space-grade FPGA in order to achieve the on-board processing of very-high resolution remotely sensed images as soon as the data are provided by the sensor. All this work has been conducted within the EU-funded Video Imaging Demonstrator for Earth Observation project. As it will be presented in this article, the work includes the introduction of a methodology based on the project constraints, the evaluation of different state-of-the-art CNN architectures by means of a new efficiency measurement also proposed in this work, the introduction of a new efficient CNN architecture, and finally, its optimized hardware implementation by means of high-level synthesis tools. The results obtained following the proposed methodology demonstrate that the uncovered architecture is able to detect targets of interest in RGB images with a much higher efficiency than state-of-the-art solutions, while requiring a much smaller amount of computing and memory resources.

1 September 2013
Verified
Conference Paper
Closed access
PCPedro P. Carballo
Pedro P. Carballo
OEOmar Espino
Omar Espino
RNRomen Neris
Romen Neris
PHPedro Hernandez-Fernandez
Pedro Hernandez-Fernandez
TSTomasz M. Szydzik
Tomasz M. Szydzik
ANAntonio Nuñez
Antonio Nuñez
This paper describes key concepts in the design and implementation of a deblocking filter (DF) for a H.264/SVC video decoder. The DF supports QCIF and CIF video formats with temporal and spatial scalability. The design flow starts from a SystemC functional model and has been refined using high-level synthesis methodology to RTL micro architecture. The process is guided with performance measurements (latency, cycle time, power, resource utilization) with the objective of assuring the quality of results of the final system. The functional model of the DF is created in an incremental way from the AVC DF model using OpenSVC source code as reference. The design flow continues with the logic synthesis and the implementation on the FPGA using various strategies. The FPGA implementation is capable to run at 100 MHz, and macro blocks are processed in 6, 500 clock cycles for a throughput of 130 fps for QCIF format and 37 fps for CIF format. A validation platform has been developed using the embedded PowerPC processor in the FPGA, composing a SoC that integrates the tasks for frame generation and visualization on a TFT screen. The FPGA implements both the DF core and a General Purpose Memory Controller (GPMC) slave core. Both cores are connected to the PowerPC440 embedded processor using Local Link interfaces. The FPGA also contains a local memory capable of storing information necessary to filter a complete frame and to store a decoded picture frame. The complete system is implemented in a Virtex5 FX70T device. An ASIC implementation of the deblocking filter has been done using UMC CMOS 65nm technology. The ASIC implementation is running at 181.8 MHz, occupying an area of 596, 392.4 μm 2 .