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.

No abstract

A closed-form and explicit 2-D analytical model for crosstalk (CTK) effects in p-n + CMOS photodiodes for pixel design optimization has been developed in this paper. This model complements and extends a previous development describing the photocurrent because of the active area illumination along with the lateral depletion region and lateral components owing to the diffused photocarriers from the surroundings of the junction. The model has very few fitting parameters because it is physically based. Similarly, it can be of great use for CMOS image sensors designers, especially to fulfill high resolution and small area requirements by pixel size reduction. The model was validated extensively through device simulations with ATLAS and experimental data, and describes the CTK dependencies on light conditions and physical, geometrical, and process parameters.

The response and crosstalk (CTK) of the p-n well photodiode were studied through device simulations performed with ATLAS and experimental data. As a result, a closed-form and explicit 2-D analytical model for its photoresponse and CTK was developed. The model has very few fitting parameters since it is physically based and describes the CTK dependencies on light conditions and physical, geometrical, and process parameters. This is of great interest for pixel design optimization to fulfill high resolution and small area requirements driven by pixel size reduction. As this model extends a previous one focused on p-n + devices, the behavior of both the structures was also compared.