DrsNet : Dual-resolution semantic segmentation with rare class-oriented superpixel prior
- PDF / 3,344,432 Bytes
- 20 Pages / 439.642 x 666.49 pts Page_size
- 81 Downloads / 206 Views
DrsNet : Dual-resolution semantic segmentation with rare class-oriented superpixel prior Liangjiang Yu1 · Guoliang Fan1 Received: 8 July 2019 / Revised: 1 August 2020 / Accepted: 21 August 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Rare-class objects in natural scene images that are usually small and less frequent often convey more important information for scene understanding than the common ones. However, they are often overlooked in scene labeling studies due to two main reasons, low occurrence frequency and limited spatial coverage. Many methods have been proposed to enhance overall semantic labeling performance, but only a few consider rare-class objects. In this work, we present a deep semantic labeling framework with special consideration of rare classes via three techniques. First, a novel dual-resolution coarse-to-fine superpixel representation is developed, where fine and coarse superpixels are applied to rare classes and background areas respectively. This unique dual representation allows seamless incorporation of shape features into integrated global and local convolutional neural network (CNN) models. Second, shape information is directly involved during the CNN feature learning for both frequent and rare classes from the re-balanced training data, and also explicitly involved in data inference. Third, the proposed framework incorporates both shape information and the CNN architecture into semantic labeling through a fusion of probabilistic multi-class likelihood. Experimental results demonstrate competitive semantic labeling performance on two standard datasets both qualitatively and quantitatively, especially for rare-class objects. Keywords Semantic segmentation · Rare class · Dual resolution · Super-pixels · Label transfer · CNN
1 Introduction The goal of scene labeling is to assign a semantic label to each pixel in an image, leading to simultaneous segmentation and recognition for scene understanding [2, 16, 22, 24, 25, 29, 30, 46, 63, 69, 72]. It is one of the most important problems of computer vision Guoliang Fan
[email protected] Liangjiang Yu [email protected] 1
School of Electrical and Computer Engineering, Oklahoma State University, Oklahoma, USA
Multimedia Tools and Applications
and pattern recognition that plays a significant role in many applications, such as robotics [44], autonomous driving [9], multimedia retrieval [69], e-commerce [32]. In most natural scene images, rare-class objects that are usually small and less frequent could be easily neglected when achieving high overall accuracy is the main goal [64]. However, rare-class objects (e.g., human, boat, vehicle, etc.) are often important for image understanding. A few deep learning approaches were recently proposed to emphasize rare-class objects, most of which encourage balanced data distribution among all classes during the training. This effect is obtained by various sampling methods [15, 50, 64], or by assigning different weights to rare classes during the trai
Data Loading...
