This paper presents PreVIous, a methodology to predict the performance of convolutional neural networks (CNNs) in terms of throughput and energy consumption on vision-enabled devices for the Internet of Things. CNNs typically constitute a massive computational load for such devices, which are characterized by scarce hardware resources to be shared among multiple concurrent tasks. Therefore, it is critical to select the optimal CNN architecture for a particular hardware platform according to prescribed application requirements. However, the zoo of CNN models is already vast and rapidly growing. To facilitate a suitable selection, we introduce a prediction framework that allows to evaluate the performance of CNNs prior to their actual implementation. The proposed methodology is based on PreVIousNet, a neural network specifically designed to build accurate per-layer performance predictive models. PreVIousNet incorporates the most usual parameters found in state-of-the-art network architectures. The resulting predictive models for inference time and energy have been tested against comprehensive characterizations of seven well-known CNN models running on two different software frameworks and two different embedded platforms. To the best of our knowledge, this is the most extensive study in the literature concerning CNN performance prediction on low-power low-cost devices. The average deviation between predictions and real measurements is remarkably low, ranging from 3% to 10%. This means state-of-the-art modeling accuracy. As an additional asset, the fine-grained a priori analysis provided by PreVIous could also be exploited by neural architecture search engines.
Many application scenarios of edge visual inference, e.g., robotics or environmental monitoring, eventually require long periods of continuous operation. In such periods, the processor temperature plays a critical role to keep a prescribed frame rate. Particularly, the heavy computational load of convolutional neural networks (CNNs) may lead to thermal throttling and hence performance degradation in few seconds. In this paper, we report and analyze the long-term performance of 80 different cases resulting from running five CNN models on four software frameworks and two operating systems without and with active cooling. This comprehensive study was conducted on a low-cost edge platform, namely Raspberry Pi 4B (RPi4B), under stable indoor conditions. The results show that hysteresis-based active cooling prevented thermal throttling in all cases, thereby improving the throughput up to approximately 90% versus no cooling. Interestingly, the range of fan usage during active cooling varied from 33% to 65%. Given the impact of the fan on the power consumption of the system as a whole, these results stress the importance of a suitable selection of CNN model and software components. To assess the performance in outdoor applications, we integrated an external temperature sensor with the RPi4B and conducted a set of experiments with no active cooling in a wide interval of ambient temperature, ranging from 22 °C to 36 °C. Variations up to 27.7% were measured with respect to the maximum throughput achieved in that interval. This demonstrates that ambient temperature is a critical parameter in case active cooling cannot be applied.
While providing the same functionality, the various Deep Learning software frameworks available these days do not provide similar performance when running the same network model on a particular hardware platform. On the contrary, we show that the different coding techniques and underlying acceleration libraries have a great impact on the instantaneous throughput and CPU utilization when carrying out the same inference with Caffe, OpenCV, TensorFlow and Caffe2 on an ARM Cortex-A53 multi-core processor. Direct modelling of this dissimilar performance is not practical, mainly because of the complexity and rapid evolution of the toolchains. Alternatively, we examine how the hardware resources are distinctly exploited by the frameworks. We demonstrate that there is a strong correlation between inference performance-including power consumption-and critical parameters associated with memory usage and instruction flow control. This identified correlation is a preliminary step for the development of a simple empirical model. The objective is to facilitate selection and further performance tuning among the ever-growing zoo of deep neural networks and frameworks, as well as the exploration of new network architectures.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.