Improving Semantic Image Segmentation via Label Fusion in Semantically Textured Meshes

TL;DR

This paper introduces an unsupervised label fusion framework using 3D meshes and semantic textures to enhance pixel-level semantic segmentation in videos, significantly improving accuracy over existing methods.

Contribution

It presents the first publicly available label fusion framework based on semantic meshes, leveraging CUDA for uncertainty-aware fusion of multi-frame predictions.

Findings

Improved pixel accuracy from 52.05% to 58.25% on ScanNet dataset.

Utilizes entire probability distribution of classes for better segmentation.

Provides open-source code to facilitate future research.

Abstract

Models for semantic segmentation require a large amount of hand-labeled training data which is costly and time-consuming to produce. For this purpose, we present a label fusion framework that is capable of improving semantic pixel labels of video sequences in an unsupervised manner. We make use of a 3D mesh representation of the environment and fuse the predictions of different frames into a consistent representation using semantic mesh textures. Rendering the semantic mesh using the original intrinsic and extrinsic camera parameters yields a set of improved semantic segmentation images. Due to our optimized CUDA implementation, we are able to exploit the entire $c$-dimensional probability distribution of annotations over $c$ classes in an uncertainty-aware manner. We evaluate our method on the Scannet dataset where we improve annotations produced by the state-of-the-art segmentation network ESANet from $52.05 \%$ to $58.25 \%$ pixel accuracy. We publish the source code of our framework online to foster future research in this area (\url{https://github.com/fferflo/semantic-meshes}). To the best of our knowledge, this is the first publicly available label fusion framework for semantic image segmentation based on meshes with semantic textures.