NTIRE 2021 NonHomogeneous Dehazing Challenge Report

Ancuti, Cosmin; Vasluianu, Florin-Alexandru; Timofte, Radu; Fu, Minghan; Liu, Huan; Yu, Yankun; Chen, Jun; Wang, Keyan; Chang, Jerome; Wang, Xiyao; Liu, Jing; Xu, Yi; Zhang, Xinjian; Zhao, Minzhi; Zhou, Shuigeng; Chen, Tianyi; Fu, Jiahui; Jiang, Wentao; Gao, Chen; Liu, Si; Wang, Yudong; Guo, Jun; Li, Chongyi; Yan, Qixin; Zheng, Sida; Zamir, Syed Waqas; Arora, Aditya; Dudhane, Akshay; Khan, Salman; Hayat, Munawar; Khan, Fahad Shahbaz; Shao, Ling; Zhang, Haichuan; Guo, Tiantong; Monga, Vishal; Yang, Wenzhu; Lin, Jingquan; Luo, Xiaotong; Huang, Guowen; Chen, Shuxin; Qu, Yanyun; Xu, Kele; Yang, Lehan; Sun, Pengliang; Niu, Xuetong; Zheng, Jun; Ruan, Xiaotong; Wang, Yunfeng; Jiang, Yang; Luo, Zhipeng; Wang, Qianqian; Xu, Zhenyu; Cao, Xiaochun; Luo, Jun; Zheng, Zhuoran; Ren, Wenqi; Wang, Tao; Chen, Yiqun; Leng, Cong; Li, Chenghua; Cheng, Jian; Sung, Chang-Sung; Chen, Jun-Cheng; Jo, Eunsung; Sim, Jae-Young; Geethu, M M; Akhil, K A; Sreeni, K. G.; Jeena, R. S.; Zacharias, Joseph; Manu, Chippy M; Huang, Zexi; Zhang, Baofeng; Zhang, Yiwen; Li, Jindong; Chen, Mianjie; Xiao, Qing; Su, Qingchao; Han, Liang; Huang, Yanting; Prajapati, Kalpesh; Chudasama, Vishal; Patel, Heena; Sarvaiya, Anjali; Raja, Kiran B.; Ramachandra, Raghavendra; Busch, Christoph; Jing, Hongyuan; Huang, Zilong; Fu, Yiran; Wu, Haoqiang; Zha, Quanxing; Zhu, Zhiwei; Lv, Hejun

doi:10.1109/cvprw53098.2021.00074

Cited by 29 publications

(8 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although there is a small haze area remaining in the third and fourth images, the color and the outline of the leaves are clearly displayed. In the challenge report of NTIRE2021 (Ancuti et al, 2021), our dehazed results achieve 21.0183 and 0.8370 in PSNR and SSIM, respectively. This method…”

Section: Results On Nh-haze2 Datasetmentioning

confidence: 87%

See 1 more Smart Citation

A Two-branch Neural Network for Non-homogeneous Dehazing via Ensemble Learning

Huan

et al. 2021

2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

View full text Add to dashboard Cite

Recently, there has been rapid and significant progress on image dehazing. Many deep learning based methods have shown their superb performance in handling homogeneous dehazing problems. However, we observe that even if a carefully designed convolutional neural network (CNN) can perform well on large-scaled dehazing benchmarks, the network usually fails on the non-homogeneous dehazing datasets introduced by NTIRE challenges. There are two main reasons. Firstly, the non-uniformly distributed haze is harder to be removed than the homogeneous haze. Secondly, the NTIRE challenge only provides limited data. To this end, we propose a simple but effective approach for non-homogeneous dehazing via ensemble learning. To be specific, we introduce a two-branch neural network to separately deal with the aforementioned problems and then map their distinct features by a learnable fusion tail. We present extensive experimental results to illustrate the effectiveness of our proposed method.

show abstract

Section: Results On Nh-haze2 Datasetmentioning

confidence: 87%

“…• We demonstrate the effectiveness of using ImageNet pre-training in the nonhomogeneous dehazing challenge (Ancuti et al, 2021).…”

Section: Hazy Oursmentioning

confidence: 99%

A Two-branch Neural Network for Non-homogeneous Dehazing via Ensemble Learning

Huan

et al. 2021

2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

View full text Add to dashboard Cite

show abstract

“…In fact, the Twobranch dehazing method significantly reduces the detection results of CasDou's detection result (−1.6% AP), while the DW-GAN is proved to be more effective when helping Cascade R-CNN improve its detection results were improved. This could be explained by the fact that DW-GAN has higher results than Two-branch dehazing when both of these two methods use the same dataset [37]. In addition, although the AP result is not so high, the DETR shows that this method can outperform all other methods by a large margin when detecting small objects, which are Pedestrian and Motor, on all three datasets (shown in Fig.…”

Section: Resultsmentioning

confidence: 91%

“…2) Two-Branch Dehazing: [18] proposed another twobranch neural network for non-homogeneous dehazing via ensemble learning. The authors found that a carefully built CNN frequently fails on a non-homogeneous dehazing dataset introduced by NITRE challenges [37] even though it performs well on large-scaled dehazing bench-marks. Therefore, they introduced a two-branch neural network to deal with the aforementioned problems separately, followed by a learnable fusion tail to map their different features.…”

Section: ) Dw Gan (Discrete Wavelet Transform Gan): Dw-gan Was Propos...mentioning

confidence: 99%

Analysis of the Influence of De-hazing Methods on Vehicle Detection in Aerial Images

Nguyen¹,

Nguyen²,

Bui³

et al. 2022

IJACSA

View full text Add to dashboard Cite

In recent years, object detection from space in adverse weather, incredibly foggy, has been challenging. In this study, we conduct an empirical experiment using two de-hazing methods: DW-GAN and Two-Branch, for removing fog, then evaluate the detection performance of six advanced object detectors belonging to four main categories: two-stage, one-stage, anchorfree and end-to-end in original and de-hazed aerial images to find the best suitable solution for vehicle detection in foggy weather. We use the UIT-DroneFog dataset, a challenging dataset that includes a lot of small, dense objects captured in various altitudes, as the benchmark to evaluate the effectiveness of approaches. After experiments, we observe that each de-hazing method has different impacts on six experimental detectors.

show abstract

“…This workshop proposes challenges in image and video processing in several fields. For instance, homogeneous [5,6] and non-homogeneous [7,8] fog removal were among the topics of interest explored for some years in the workshop. In these challenges, some research groups exploited previous information on the image and tried to evaluate the natural parameters through deep learning techniques [9,10].…”

Section: Introductionmentioning

confidence: 99%

Contour-Based Metric for the Evaluation of Image Defogging without a Ground Truth

DeMas-Giménez¹,

García-Gómez²,

Casas³

et al. 2022

Preprint

View full text Add to dashboard Cite

Fog, haze, or smoke are usual atmospheric phenomena that dramatically compromise the overall visibility of any scene, critically affecting features such as illumination, contrast, and contour detection of objects. The decrease in visibility compromises the performance of computer vision algorithms such as pattern recognition and segmentation, some of them very relevant for decision-making for the security or autonomous vehicle industries. Several dehazing methods have been proposed, however, to the best of our knowledge, all proposed metrics compare the defogged image to its ground truth for evaluation of the defogging algorithms, or need to estimate parameters through physical models. This fact hinders progress in the field as obtaining proper ground truth images is costly and time-consuming, and physical parameters greatly depend on the scene conditions. This paper aims to tackle this issue by proposing a contour-based metric for image defogging evaluation that does not need a ground truth image. The proposed metric only requires the original hazy RGB image and the RGB image after the defogging procedure. A comparison of the proposed metric with metrics currently used in the NTIRE 2018 defogging challenge is performed to prove its effectiveness in a general situation, showing comparable results to conventional metrics.

show abstract

NTIRE 2021 NonHomogeneous Dehazing Challenge Report

Cited by 29 publications

References 38 publications

A Two-branch Neural Network for Non-homogeneous Dehazing via Ensemble Learning

A Two-branch Neural Network for Non-homogeneous Dehazing via Ensemble Learning

Analysis of the Influence of De-hazing Methods on Vehicle Detection in Aerial Images

Contour-Based Metric for the Evaluation of Image Defogging without a Ground Truth

Contact Info

Product

Resources

About