Unsupervised Day–Night Domain Adaptive Image Classification Based on Target-Domain Discriminative Enhancement
Open Access
- 22 September 2026
- journal article
- research article
- Published by MDPI AG in Applied Sciences
- Vol. 16 (19) , 9395
Abstract
Day–night image classification under low-light conditions faces substantial domain shift, uncertain target-domain predictions, and unreliable pseudo-label supervision. To address these challenges, this paper proposes an unsupervised day–night domain adaptation method for image classification based on target-domain discriminative enhancement. Built on a bidirectional day–night domain alignment framework, the method introduces Target-Domain Discriminative Clustering (TDCL) and Confidence-Calibrated Pseudo-Label Learning (ConCal). TDCL combines sample-level entropy minimization with batch-level nuclear-norm maximization to improve prediction certainty and the discriminative structure of target-domain predictions. ConCal applies class-adaptive confidence thresholds to current-model predictions to select reliable target samples and uses the corresponding pseudo-labels for self-training. On the CODaN nighttime test set, the proposed method achieves an accuracy of 76.24 ± 0.41%, exceeding the published WiiD result by 5.92 percentage points. On the extended ExDark evaluation set, comprising additional non-overlapping nighttime images, it achieves 76.73 ± 0.57%, outperforming the reproduced WiiD baseline by 3.63 percentage points. These results indicate that explicitly optimizing the target-domain prediction structure and improving pseudo-label reliability can further enhance nighttime classification.Keywords
This publication has 42 references indexed in Scilit:
- Progressive Dual-Enhancement and Auxiliary Domain Guidance for Low-Light Domain Adaptive Object DetectionSignal, Image and Video Processing, 2026
- Cascade-UDA: A Cascade paradigm for unsupervised domain adaptationNeurocomputing, 2025
- H3T: Hierarchical Transferable Transformer with TokenMix for Unsupervised Domain AdaptationExpert Systems with Applications, 2024
- Uncertainty-aware pseudo-label filtering for source-free unsupervised domain adaptationNeurocomputing, 2023
- Unsupervised domain adaptation via distilled discriminative clusteringPattern Recognition, 2022
- Learning to Enhance Low-Light Image via Zero-Reference Deep Curve EstimationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2021
- EnlightenGAN: Deep Light Enhancement Without Paired SupervisionIEEE Transactions on Image Processing, 2021
- Getting to know low-light images with the Exclusively Dark datasetComputer Vision and Image Understanding, 2019
- Adaptive Batch Normalization for practical domain adaptationPattern Recognition, 2018
- The Retinex Theory of Color VisionScientific American, 1977