Brain-Inspired Classroom Occupancy Monitoring on a Low-Power Mobile Platform

Conti, Francesco; Pullini, Antonio; Benini, Luca

doi:10.1109/cvprw.2014.95

Cited by 29 publications

(26 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, in this case (as is shown in Figure 5) we consider a very small CNN architecture that begins with a strong reduction in the dimensionality of the input (using a 4:1 max-pooling layer) to reduce the computational complexity of the model. Our CNN implementation is based the CConvNet library [43], that takes advantage of the OpenMP programming model for better performance on the parallel PULP platform.…”

Section: B Visual Feature Extraction On Pulpmentioning

confidence: 99%

Accelerated Visual Context Classification on a Low-Power Smartwatch

Conti

Palossi

Andri

et al. 2016

IEEE Trans. Human-Mach. Syst.

Self Cite

View full text Add to dashboard Cite

Data produced by wearable sensors is key in contexts such as performance enhancement and training help for sports and fitness, continuous monitoring for aging people and for chronic disease management, and in gaming and entertainment. Unfortunately, wearable devices currently on the market are either incapable of complex functionality or severely impaired by short battery lifetime. In this work, we present a smartwatch platform based on an ultra-low power (ULP) heterogeneous system composed by a TI MSP430 microcontroller, the PULP programmable parallel accelerator and a set of ULP sensors, including a camera. The embedded PULP accelerator enables state-of-the-art context classification based on Convolutional Neural Networks (CNNs) to be applied within a sub-10mW system power envelope. Our methodology enables to reach high accuracy in context classification over 5 classes (up to 84%, with 3 classes over 5 reaching more than 90% accuracy), while consuming 2.2mJ per classification, or an ultra-low energy consumption of less than 91uJ per classification with an accuracy of 64%-3.2× better than chance. Our results suggest that the proposed heterogeneous platform can provide up to 500× speedup with respect to the MSP430 within a similar power envelope, which would enable complex computer vision algorithms to be executed in highly power-constrained scenarios.

show abstract

Section: B Visual Feature Extraction On Pulpmentioning

confidence: 99%

Accelerated Visual Context Classification on a Low-Power Smartwatch

Conti

Palossi

Andri

et al. 2016

IEEE Trans. Human-Mach. Syst.

Self Cite

View full text Add to dashboard Cite

show abstract

“…For our experiments we consider six applications that were originally developed for the homogeneous P2012 cluster: Viola-Jones Face Detection using Haar features [48], FAST Circular Corner Detection [38,39], Removed Object Detection based on normalized cross-correlation (NCC) [26], a parallel version of Color Tracking from the OpenCV library [3], a Convolutional Neural Network (CNN) forward-propagation benchmark [13,25], and a Mahalanobis Distance (10-dimensional) kernel. Face Detection is parallelized with OpenCL, while the other benchmarks use the OpenMP implementation described in Marongiu et al [27].…”

Section: Resultsmentioning

confidence: 99%

He-P2012: Performance and Energy Exploration of Architecturally Heterogeneous Many-Cores

Conti

Marongiu

Pilkington

et al. 2015

J Sign Process Syst

Self Cite

View full text Add to dashboard Cite

The end of Dennardian scaling in advanced technologies brought about new architectural templates to overcome the so-called utilization wall and provide Moore's Law-like performance and energy scaling in embedded SoCs. One of the most promising templates, architectural heterogeneity, is hindered by high cost due to the design space explosion and the lack of effective exploration tools. Our work provides three contributions towards a scalable and effective methodology for design space exploration in embedded MC-SoCs. First, we present the He-P2012 architecture, augmenting the state-of-art STMicroelectronics P2012 platform with heterogeneous shared-L1 coprocessors called HW processing elements (HWPE). Second, we propose a novel methodology for the semi-automatic definition and instantiation of shared-memory HWPEs from a C source, supporting both simple and structured data types. Third, we demonstrate that the integration of HWPEs can provide significant performance and energy efficiency benefits on a set of benchmarks originally developed for the homogeneous P2012, achieving up to 123x speedup on the accelerated code region (∼98 % of Amdahl's law limit) while saving 2/3 of the energy.

show abstract

“…The images were split into different numbers of blocks, with one PE working on one block. • Using convolutional neural networks taken from the CConvNet library [7,8], up to four hand-written digits were simultaneously recognised out of images, with four PEs working together for one recognition. • Floating point number arrays of different sizes were sorted using a merge sort algorithm.…”

Section: Performance Comparison Of Interleaved Vs Contiguous Mappingmentioning

confidence: 99%

Minimising Access Conflicts on Shared Multi-Bank Memory

Tretter

Giannopoulou

Baer

et al. 2017

ACM Trans. Embed. Comput. Syst.

View full text Add to dashboard Cite

A common multi-core pattern consists of processors communicating through shared, multi-banked on-chip memory. Two approaches exist: Interleaved address mapping, which spreads consecutive data over all banks, and contiguous address mapping, which stores consecutive data on a single bank. In this work, we compare both approaches on the Kalray MPPA-256 platform. For contiguous mapping, we propose an algorithm, based on graph colouring techniques, to automatically perform the assignment of data blocks to memory banks with the goal of minimising access collisions and delays. Experiments with representative, parallel real-world benchmarks show that 69% of the tested configurations, when optimised for contiguous mapping by our algorithm, run up to 86% faster on average than with interleaved mapping. CCS Concepts: • Computing methodologies → Shared memory algorithms; • Computer systems organization → Embedded software; System on a chip; • Hardware → Memory and dense storage;

show abstract

Brain-Inspired Classroom Occupancy Monitoring on a Low-Power Mobile Platform

Cited by 29 publications

References 20 publications

Accelerated Visual Context Classification on a Low-Power Smartwatch

Accelerated Visual Context Classification on a Low-Power Smartwatch

He-P2012: Performance and Energy Exploration of Architecturally Heterogeneous Many-Cores

Minimising Access Conflicts on Shared Multi-Bank Memory

Contact Info

Product

Resources

About