DFNet: Semantic Segmentation on Panoramic Images with Dynamic Loss Weights and Residual Fusion Block

TL;DR

This paper introduces DFNet, a novel semantic segmentation approach for panoramic images in autonomous driving, utilizing dynamic loss weights and a residual fusion block to improve accuracy.

Contribution

The paper presents two innovations: dynamic loss weights based on class pixel count and a residual fusion block for enhanced feature integration.

Findings

Achieved state-of-the-art results on PSV dataset

Improved segmentation accuracy for lane markings and parking slots

Demonstrated effectiveness of RFB in feature fusion

Abstract

For the self-driving and automatic parking, perception is the basic and critical technique, moreover, the detection of lane markings and parking slots is an important part of visual perception. In this paper, we use the semantic segmentation method to segment the area and classify the class of lane makings and parking slots on panoramic surround view (PSV) dataset. We propose the DFNet and make two main contributions, one is dynamic loss weights, and the other is residual fusion block (RFB). Dynamic loss weights are varying from classes, calculated according to the pixel number of each class in a batch. RFB is composed of two convolutional layers, one pooling layer, and a fusion layer to combine the feature maps by pixel multiplication. We evaluate our method on PSV dataset, and achieve an advanced result.