Edge Computing for Vision-Based, Urban-Insects Traps in the Context of Smart Cities

Abstract

Our aim is to promote the widespread use of electronic insect traps that report captured pests to a human-controlled agency. This work reports on edge-computing as applied to camera-based insect traps. We present a low-cost device with high power autonomy and an adequate picture quality that reports an internal image of the trap to a server and counts the insects it contains based on quantized and embedded deep-learning models. The paper compares different aspects of performance of three different edge devices, namely ESP32, Raspberry Pi Model 4 (RPi), and Google Coral, running a deep learning framework (TensorFlow Lite). All edge devices were able to process images and report accuracy in counting exceeding 95%, but at different rates and power consumption. Our findings suggest that ESP32 appears to be the best choice in the context of this application according to our policy for low-cost devices.

Codling Moth Monitoring with Camera-Equipped Automated Traps: A Review

“…Those factors make most of the earlier developed insect classifier methods impractical where images were of high quality and finely segmented. Therefore, image counting is a more complex problem, and the literature is sparse because the main research trend is the identification [54].…”

“…Therefore, deep object detectors can be easily ported. Although there are some microcontroller boards such as the ESP32 that supports the adaptation of deep artificial neural network models due to the Tensorfow Lite library [54], the restricted development environment and computational resources do not allow the adaptation of the widely used deep detectors. Finally, the weaknesses of single board computers can be eliminated by external circuit.…”

An Adaptable and Unsupervised TinyML Anomaly Detection System for Extreme Industrial Environments

“…This system is based on Raspberry Pi and Arduino boards embedding a YOLOv4-tiny and SSD-MobileNet deep networks. A similar approach was proposed by Saradopoulos et al [ 21 ] that trains a DNN with a properly crafted dataset and then compares the system performance of their different edge devices (i.e., ESP32, Raspberry PI 4, and Google Coral Dev Board) in term of power consumption, accuracy, the processing time, and memory footprint. In the intelligent traffic management domain, Zhang et al [ 22 ] adapt the YOLOv5 network to detect zebra-crossing using images captured by a camera mounted in the front of a car.…”

Image-Based Insect Counting Embedded in E-traps that Learn Without Manual Image Annotation and Self-Dispose Captured Insects

“…Therefore, it outputs a single number of insects counts without generating bounding boxes or identifying species in the process. Models of this kind are lighter than the other methods and embeddable to microprocessors (see [27]). The training of the network is performed in forward and backward stages based on the prediction output and the labelled ground-truth as provided by the image synthesis stage.…”

“…The transmission of the images introduces a large bandwidth overhead that raises communications costs and power consumption and can compromise the design of the system that must use low-quality picture analysis to mitigate these costs. Therefore, the current research trend -where also our work belongs-is to embed sophisticated deeplearning based (DL) systems in the device deployed in the field (edge computing) and transmit only the results (i.e., counts of insects, environmental variables such as ambient humidity and temperature, GPS coordinates and timestamps) [27][28]. Moreover, such a low-data approach allows for a network of LoRa based nodes with a common gateway that uploads the data, further reducing communication costs.…”