Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deformable Convolution Based Road Scene Semantic Segmentation of Fisheye Images in Autonomous Driving

Anam Manzoor, Aryan Singh, Ganesh Sistu, Reenu Mohandas, Eoin Grua, Anthony Scanlan, Ciarán Eising

TL;DR

The significant improvement in mIoU score resulting from integrating Deformable CNNs demonstrates their effectiveness in handling the geometric distortions present in fisheye imagery, exceeding the performance of traditional CNN architectures.

Abstract

This study investigates the effectiveness of modern Deformable Convolutional Neural Networks (DCNNs) for semantic segmentation tasks, particularly in autonomous driving scenarios with fisheye images. These images, providing a wide field of view, pose unique challenges for extracting spatial and geometric information due to dynamic changes in object attributes. Our experiments focus on segmenting the WoodScape fisheye image dataset into ten distinct classes, assessing the Deformable Networks' ability to capture intricate spatial relationships and improve segmentation accuracy. Additionally, we explore different loss functions to address class imbalance issues and compare the performance of conventional CNN architectures with Deformable Convolution-based CNNs, including Vanilla U-Net and Residual U-Net architectures. The significant improvement in mIoU score resulting from integrating Deformable CNNs demonstrates their effectiveness in handling the geometric distortions present in fisheye imagery, exceeding the performance of traditional CNN architectures. This underscores the significant role of Deformable convolution in enhancing semantic segmentation performance for fisheye imagery.

Deformable Convolution Based Road Scene Semantic Segmentation of Fisheye Images in Autonomous Driving

TL;DR

The significant improvement in mIoU score resulting from integrating Deformable CNNs demonstrates their effectiveness in handling the geometric distortions present in fisheye imagery, exceeding the performance of traditional CNN architectures.

Abstract

This study investigates the effectiveness of modern Deformable Convolutional Neural Networks (DCNNs) for semantic segmentation tasks, particularly in autonomous driving scenarios with fisheye images. These images, providing a wide field of view, pose unique challenges for extracting spatial and geometric information due to dynamic changes in object attributes. Our experiments focus on segmenting the WoodScape fisheye image dataset into ten distinct classes, assessing the Deformable Networks' ability to capture intricate spatial relationships and improve segmentation accuracy. Additionally, we explore different loss functions to address class imbalance issues and compare the performance of conventional CNN architectures with Deformable Convolution-based CNNs, including Vanilla U-Net and Residual U-Net architectures. The significant improvement in mIoU score resulting from integrating Deformable CNNs demonstrates their effectiveness in handling the geometric distortions present in fisheye imagery, exceeding the performance of traditional CNN architectures. This underscores the significant role of Deformable convolution in enhancing semantic segmentation performance for fisheye imagery.
Paper Structure (7 sections, 4 figures, 1 table)

This paper contains 7 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background Study on Semantic Segmentation in Automotive Imaging
  3. Methodology
  4. Dataset
  5. Experimental Setup
  6. Results and Experiments
  7. Conclusion and Future Work

Figures (4)

  • Figure 1: Four fisheye cameras mounted around the vehicle to provide complete 360-degree coverage.
  • Figure 1: (a)
  • Figure 2: Baseline Vanilla DeU-Net model where Deformable Convolution block injected into the first layer of the encoder and last layer of decoder path to better account the spatial and geometric characteristics of fisheye images during training.
  • Figure 4: Visualizations of results on Woodscape fisheye images and corresponding ground truth masks are presented across baseline models including Vanilla_U-Net, Residual_U-Net, Deformable_U-Net, and Deformable_Residual model. Notably, the visualization performance with Vanilla_DeU-Net surpasses the other models, as indicated with green boxes compared to ground truth masks, with particular emphasis on distorted edges.