BiMaL: Bijective Maximum Likelihood Approach to Domain Adaptation in\n Semantic Scene Segmentation

2021/08/06 by Thanh-Dat Truong, Chi Nhan Duong, Truong, Thanh-Dat +9 · 4 citations
Computer Science · Medicine · #Domain Adaptation and Few-Shot Learning #Advanced Neural Network Applications #COVID-19 diagnosis using AI

paper · pdf · doi:10.48550/arxiv.2108.03267

Abstract

Semantic segmentation aims to predict pixel-level labels. It has become a\npopular task in various computer vision applications. While fully supervised\nsegmentation methods have achieved high accuracy on large-scale vision\ndatasets, they are unable to generalize on a new test environment or a new\ndomain well. In this work, we first introduce a new Un-aligned Domain Score to\nmeasure the efficiency of a learned model on a new target domain in\nunsupervised manner. Then, we present the new Bijective Maximum\nLikelihood(BiMaL) loss that is a generalized form of the Adversarial Entropy\nMinimization without any assumption about pixel independence. We have evaluated\nthe proposed BiMaL on two domains. The proposed BiMaL approach consistently\noutperforms the SOTA methods on empirical experiments on "SYNTHIA to\nCityscapes", "GTA5 to Cityscapes", and "SYNTHIA to Vistas".\n

Cited by

• CLIFF: Continual Learning for Incremental Flake Features in 2D Material Identification
• MANGO: Multimodal Attention-based Normalizing Flow Approach to Fusion Learning
• φ-Adapt: A Physics-Informed Adaptation Learning Approach to 2D Quantum Material Discovery
• BIMA: Bijective Maximum Likelihood Learning Approach to Hallucination Prediction and Mitigation in Large Vision-Language Models

Related

• ADVENT: Adversarial Entropy Minimization for Domain Adaptation in\n Semantic Segmentation
• Alleviating Semantic-level Shift: A Semi-supervised Domain Adaptation Method for Semantic Segmentation
• Threshold-adaptive Unsupervised Focal Loss for Domain Adaptation of Semantic Segmentation
• PixMatch: Unsupervised Domain Adaptation via Pixelwise Consistency\n Training
• Bidirectional Learning for Domain Adaptation of Semantic Segmentation