Exploiting Optimized Depthwise Separable Convolutions for Traffic Signal Recognition

Traffic signal recognition using Optimized Depthwise Separable Convolutions (ODSCs) is more proficient than established Convolutional Neural Networks (CNNs) for ongoing projects such as self-drive car and intelligent transportation systems. Efficiency of Optimized Depthwise Separable Convolutions is improved in two modest steps are depthwise convolutions and pointwise convolutions, these operation uses effectively utilize the memory, reduces the processing time and supports scalability. Prediction of traffic sign using ODSCs improves accuracy rate compare to CNNs and also increases the speed of the computations, consumption of energy is less and reduced model size. ODSCs are well suited for device with minimum resource such as embedded system in transportation, intelligent city infrastructures to recognition of traffic signals.