Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..

Towards Unsupervised Domain Bridging via Image Degradation in Semantic Segmentation

Authors: Wangkai Li, Rui Sun, Huayu Mai, Tianzhu Zhang

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on multiple domain adaptive semantic segmentation benchmarks demonstrate that Di DA consistently achieves significant performance improvements across all settings.
Researcher Affiliation Academia 1Mo E Key Laboratory of Brain-inspired Intelligent Perception and Cognition, University of Science and Technology of China 2National Key Laboratory of Deep Space Exploration, Deep Space Exploration Laboratory EMAIL, EMAIL
Pseudocode Yes Algorithm 1 Pseudo algorithms of Di DA.
Open Source Code Yes Code is available at https://github.com/Woof6/Di DA.
Open Datasets Yes As synthetic datasets, we use GTAv [70] containing 24,966 images and SYNTHIA [71] with 9,400 images. For real-world datasets, we employ Cityscapes [21] with 2,975 training and 500 validation images representing clear weather conditions, and ACDC [72] containing 1,600 training, 406 validation, and 2,000 test images capturing adverse weather conditions (fog, night, rain, and snow).
Dataset Splits Yes Cityscapes [21] with 2,975 training and 500 validation images representing clear weather conditions, and ACDC [72] containing 1,600 training, 406 validation, and 2,000 test images capturing adverse weather conditions (fog, night, rain, and snow).
Hardware Specification Yes all experiments are conducted on one or two RTX-3090 GPUs with 24 GB memory
Software Dependencies No We implement Di DA based on the MMSegmentation [20] framework.
Experiment Setup Yes all experiments are conducted on one or two RTX-3090 GPUs with 24 GB memory, with 40K training iterations and a batch size of 2. We train the network using the Adam W optimizer, with learning rates of 6 10 5 for the encoder and 6 10 4 for the decoder, a weight decay of 0.01, and a linear learning rate warm-up strategy for the first 1.5K iterations. The EMA coefficient for updating the teacher network is set to 0.999. We set T = 100