Implementation of Lightweight Convolutional Neural Networks with an Early Exit Mechanism Utilizing 40 nm CMOS Process for Fire Detection in Unmanned Aerial Vehicles

Yu-Pei Liang,Chen-Ming Chang,Ching-Che Chung

DOI: https://doi.org/10.3390/s24072265

IF: 3.9

2024-04-03

Sensors

Abstract:The advancement of unmanned aerial vehicles (UAVs) enables early detection of numerous disasters. Efforts have been made to automate the monitoring of data from UAVs, with machine learning methods recently attracting significant interest. These solutions often face challenges with high computational costs and energy usage. Conventionally, data from UAVs are processed using cloud computing, where they are sent to the cloud for analysis. However, this method might not meet the real-time needs of disaster relief scenarios. In contrast, edge computing provides real-time processing at the site but still struggles with computational and energy efficiency issues. To overcome these obstacles and enhance resource utilization, this paper presents a convolutional neural network (CNN) model with an early exit mechanism designed for fire detection in UAVs. This model is implemented using TSMC 40 nm CMOS technology, which aids in hardware acceleration. Notably, the neural network has a modest parameter count of 11.2 k. In the hardware computation part, the CNN circuit completes fire detection in approximately 230,000 cycles. Power-gating techniques are also used to turn off inactive memory, contributing to reduced power consumption. The experimental results show that this neural network reaches a maximum accuracy of 81.49% in the hardware implementation stage. After automatic layout and routing, the CNN hardware accelerator can operate at 300 MHz, consuming 117 mW of power.

engineering, electrical & electronic,chemistry, analytical,instruments & instrumentation

What problem does this paper attempt to address?

The paper attempts to address the problem of achieving efficient fire detection in Unmanned Aerial Vehicles (UAVs). Specifically, the authors focus on how to deploy lightweight Convolutional Neural Networks (CNNs) on UAVs to achieve real-time, low-power fire detection. Traditional methods often rely on sending data to the cloud for processing, which not only introduces network latency and data security issues but also fails to meet real-time requirements. While edge computing can achieve real-time processing, it still faces challenges in terms of computation and energy consumption.

To overcome these obstacles, the paper proposes a lightweight CNN model combined with an early exit mechanism and uses 40-nanometer CMOS technology for hardware acceleration. The model has the following features:

1. **Lightweight Model**: The model has only 11.2K parameters, significantly reducing computational and storage requirements.
2. **Early Exit Mechanism**: By incorporating multiple early exit points in the model, it allows flexible selection of exit points in different usage scenarios, thereby reducing computation time and energy consumption while ensuring a certain level of accuracy.
3. **Hardware Optimization**: The use of 40-nanometer CMOS technology for hardware acceleration and the adoption of power gating techniques to shut down inactive memory further reduce power consumption.
4. **High Efficiency and Accuracy**: Experimental results show that the model achieves a maximum accuracy of 81.49% during the hardware implementation phase. Additionally, after automatic placement and routing, the hardware accelerator can operate at a frequency of 300 MHz with a power consumption of only 117 mW.

In summary, the paper aims to solve the real-time and energy efficiency issues in UAV fire detection by designing and implementing a lightweight and efficient CNN model.