ComBiNet: Compact Convolutional Bayesian Neural Network for Image Segmentation

TL;DR

ComBiNet is a compact Bayesian convolutional neural network designed for image segmentation that offers uncertainty quantification and reduced resource requirements, making it more practical for real-world applications.

Contribution

It introduces a parameter-efficient Bayesian architecture combining separable convolutions and Monte Carlo Dropout for uncertainty estimation, with fewer than 2.5 million parameters.

Findings

Achieves marginal accuracy improvement over related models

Uses significantly fewer parameters and compute resources

Provides per-pixel uncertainty quantification

Abstract

Fully convolutional U-shaped neural networks have largely been the dominant approach for pixel-wise image segmentation. In this work, we tackle two defects that hinder their deployment in real-world applications: 1) Predictions lack uncertainty quantification that may be crucial to many decision-making systems; 2) Large memory storage and computational consumption demanding extensive hardware resources. To address these issues and improve their practicality we demonstrate a few-parameter compact Bayesian convolutional architecture, that achieves a marginal improvement in accuracy in comparison to related work using significantly fewer parameters and compute operations. The architecture combines parameter-efficient operations such as separable convolutions, bilinear interpolation, multi-scale feature propagation and Bayesian inference for per-pixel uncertainty quantification through Monte Carlo Dropout. The best performing configurations required fewer than 2.5 million parameters on diverse challenging datasets with few observations.