Enhancing Nighttime Semantic Segmentation with Visual-Linguistic Priors and Wavelet Transform

Authors: Jianhou Zhou, Xiaolong Zhou, Sixian Chan, Zhaomin Chen, Xiaoqin Zhang

IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on benchmarks including Night City, Night Cityfine, BDD100K, and City Scapes demonstrate our method s superior performance over existing approaches. Our contributions are summarized as follows: Extensive experiments on various challenging benchmarks show that the proposed method outperforms stateof-the-art nighttime semantic segmentation.
Researcher Affiliation Academia Jianhou Zhou1 , Xiaolong Zhou2 , Sixian Chan3 and Zhaomin Chen4 , Xiaoqin Zhang3 1Hangzhou Dianzi University 2Quzhou University 3Zhejiang University Of Technology 4Wenzhou University naiive EMAIL, EMAIL, EMAIL, EMAIL, EMAIL
Pseudocode No The paper includes figures illustrating the architecture and components (Figure 2, 3, 4) and mathematical equations, but it does not contain explicit pseudocode blocks or algorithm listings.
Open Source Code No The paper does not contain any explicit statement about releasing source code for the methodology described, nor does it provide a link to a code repository.
Open Datasets Yes Following previous practices[Deng et al., 2022; Wei et al., 2023; Cheng et al., 2022], we evaluate the nighttime semantic segmentation performance of our method on four datasets: Night City, Night City-fine, City Scapes, and BDD100K. Night City [Tan et al., 2021], the largest nighttime semantic segmentation dataset, contains 4,297 real nighttime images... City Scapes [Cordts et al., 2016], an autonomous driving dataset... BDD100K [Yu et al., 2020] is a large-scale driving dataset...
Dataset Splits Yes Night City [Tan et al., 2021], the largest nighttime semantic segmentation dataset, contains 4,297 real nighttime images, with 2,998 for training and 1,299 for validation. All images are 1024 512 in resolution... City Scapes [Cordts et al., 2016], an autonomous driving dataset, contains daytime images from 50 cities, with 2,975 training and 500 validation images, all at 2048 1024 resolution... We use a subset, BDD100K-night, for supplementary experiments, with 314 nighttime images for training and 31 for validation.
Hardware Specification No The paper mentions implementing the model using the MMSegmentation framework and using a batch size of 16, but it does not provide any specific details about the hardware (e.g., GPU models, CPU types) used for the experiments.
Software Dependencies No Our model is implemented using the MMSegmentation framework. The paper does not provide specific version numbers for MMSegmentation or any other software libraries or dependencies.
Experiment Setup Yes During training, we apply the recommended preprocessing methods from MMSegmentation, including mean normalization, random scaling, and flipping, on the Night City, Night City-fine, and City Scapes datasets. We use the default City Scapes 90k training configuration with a batch size of 16. For wavelet image reconstruction, λ = 0.1 and µ = 10, with their effects on segmentation accuracy tested in the ablation experiments. To handle size variations during inference, we use input image versions rescaled by factors of [0.5, 0.75, 1.0, 1.25, 1.5, 1.75]. Additionally, we apply horizontal flipping and average the predictions from all augmented versions.