Performance Evaluation of Deep Learning Networks for Semantic Segmentation of Traffic Stereo-Pair Images

TL;DR

This study evaluates deep learning models PSPNet and ICNet for semantic segmentation of traffic stereo images, analyzing accuracy, inference time, and statistical properties across different cities and channels, revealing their sensitivities and potential for enhanced scene understanding.

Contribution

It provides a comparative analysis of PSPNet and ICNet on traffic stereo images, highlighting their statistical distribution properties and implications for model fine-tuning and multi-channel data utilization.

Findings

Different accuracy distributions for various cities and stereo channels.

PSPNet shows more asymmetric and outlier-prone accuracy distributions than ICNet.

Stereo pairs can be used as additional data channels for scene analysis.

Abstract

Semantic image segmentation is one the most demanding task, especially for analysis of traffic conditions for self-driving cars. Here the results of application of several deep learning architectures (PSPNet and ICNet) for semantic image segmentation of traffic stereo-pair images are presented. The images from Cityscapes dataset and custom urban images were analyzed as to the segmentation accuracy and image inference time. For the models pre-trained on Cityscapes dataset, the inference time was equal in the limits of standard deviation, but the segmentation accuracy was different for various cities and stereo channels even. The distributions of accuracy (mean intersection over union - mIoU) values for each city and channel are asymmetric, long-tailed, and have many extreme outliers, especially for PSPNet network in comparison to ICNet network. Some statistical properties of these distributions (skewness, kurtosis) allow us to distinguish these two networks and open the question about relations between architecture of deep learning networks and statistical distribution of the predicted results (mIoU here). The results obtained demonstrated the different sensitivity of these networks to: (1) the local street view peculiarities in different cities that should be taken into account during the targeted fine tuning the models before their practical applications, (2) the right and left data channels in stereo-pairs. For both networks, the difference in the predicted results (mIoU here) for the right and left data channels in stereo-pairs is out of the limits of statistical error in relation to mIoU values. It means that the traffic stereo pairs can be effectively used not only for depth calculations (as it is usually used), but also as an additional data channel that can provide much more information about scene objects than simple duplication of the same street view images.