Performance Evaluation of Deep Learning Algorithm Using High-End Media Processing Board in Real-Time Environment

Table 1

A brief overview of various deep learning algorithms implemented over GPUs or conventional computing devices for a variety of object detection purposes.

Deep learning algorithms implemented over conventional computing devices
Author and publication year	Hardware	Algorithm	Aim
Artamonov et al. [22]	NVIDIA Jetson/Tegra	YOLO- CNN	Traffic sign recognition
Barba-Guaman et al. [26]	Jetson Nano	SSD-Mobilenet V1 and V2 (single shot detector), SSD-inception V2, and PedNet, multiped	Vehicle and pedestrian detection
Komasilovs et al. [12]	Intel i5, 16 GB RAM	SSD Mobilenet V1 model	Traffic sign recognition
Castellano et al. [27]	NVIDIA GeForce MX110 (2 GB), RPi3 NVIDIA Jetson TX2	Lightweight FCN	Crowd detection
Zhao et al. [6]	LiDAR(light detection and ranging) sensors, NVIDIA GTX 1080i GPU	Spiking convolutional neural network in YOLOv2	Vehicle and pedestrian detection and minimize power consumption of LiDAR
Avramović et al. [25]	GeForce GTX 1080 Ti	YOLO variants	Traffic sign recognition
Khazukov et al. [24]	GPU: GeForce RTX 2080 TI, CPU: i9 9900k, RAM: 64 GB	YOLOv3	Speed detection and classification
Komasilovs et al. [12]	Cameras, CPU Intel i5, 16 GB RAM	SSD Mobilenet V1 model	Object detection and tracking
Blair and Robertson [21]	FPGA(field programmable gate array), GPU, and CPU	HOG, MoG (Histogram of Oriented Gradient and Mixture of Gaussian)	Object tracking/event detection