Compute and memory demands of state-of-the-art deep learning methods are still a shortcoming that must be addressed to make them useful at IoT end-nodes. In particular, recent results depict a hopeful prospect for image processing using Convolutional Neural Netwoks, CNNs, but the gap between software and hardware implementations is already considerable for IoT and mobile edge computing applications due to their high power consumption. This proposal performs low-power and real time deep learning-based multiple object visual tracking implemented on an NVIDIA Jetson TX2 development kit. It includes a camera and wireless connection capability and it is battery powered for mobile and outdoor applications. A collection of representative sequences captured with the onboard camera, dETRUSC video dataset, is used to exemplify the performance of the proposed algorithm and to facilitate benchmarking. The results in terms of power consumption and frame rate demonstrate the feasibility of deep learning algorithms on embedded platforms although more effort to joint algorithm and hardware design of CNNs is needed.

Wireless Sensor Networks (WSN) are increasingly adopted in agriculture to monitor environmental variables to predict the presence of pests. Differently from these approaches, our work introduces a WSN to detect the presence of snails in the field. The network can be used to both trigger an alarm of early pest presence and to further elaborate statistical models with the addition of environmental data as temperature or humidity to predict snail presence. In this work we also design our own WSN simulator to account for real-life conditions as an uneven spacing of motes in the field or different currents generated by solar cells at the motes. This allows to achieve a more realistic network deployment in the field. Experimental tests are included in this paper, showing that our motes are perpetual in terms of energy consumption.

No abstract