4,091 Hits in 4.4 sec

High-Resolution Daytime Translation Without Domain Labels [article]

Ivan Anokhin, Pavel Solovev, Denis Korzhenkov, Alexey Kharlamov, Taras Khakhulin, Alexey Silvestrov, Sergey Nikolenko, Victor Lempitsky, Gleb Sterkin
2020 arXiv   pre-print
We present the high-resolution daytime translation (HiDT) model for this task.  ...  HiDT combines a generative image-to-image model and a new upsampling scheme that allows to apply image translation at high resolution.  ...  High-resolution translation.  ... 
arXiv:2003.08791v2 fatcat:mt4plhbpnne4vkio7jmqafcyyq

High-Resolution Daytime Translation Without Domain Labels

Ivan Anokhin, Pavel Solovev, Denis Korzhenkov, Alexey Kharlamov, Taras Khakhulin, Aleksei Silvestrov, Sergey Nikolenko, Victor Lempitsky, Gleb Sterkin
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
Figure 1: Daytime translation results. Left -original images, right -translated and enhanced images (one style per column).  ...  High-resolution translation.  ...  The main Section 3 presents the High-Resolution Daytime Translation (HiDT) model and the resolution-increasing enhancement model.  ... 
doi:10.1109/cvpr42600.2020.00751 dblp:conf/cvpr/AnokhinSKKKSNLS20 fatcat:pze2im5rkvdxrath7bkc7fr7ve

See Clearer at Night: Towards Robust Nighttime Semantic Segmentation through Day-Night Image Conversion [article]

Lei Sun, Kaiwei Wang, Kailun Yang, Kaite Xiang
2019 arXiv   pre-print
In another method, we use GANs to translate different ratio of daytime images in the dataset to the nighttime but still with their labels.  ...  In the first method, GANs were used to translate nighttime images to the daytime, thus semantic segmentation can be performed using robust models already trained on daytime datasets.  ...  to translate image across daytime and nighttime domains.  ... 
arXiv:1908.05868v1 fatcat:2bksxp2xxbfvxefnkfrm32geca

An Unsupervised Transfer Learning Framework for Visible-Thermal Pedestrian Detection

Chengjin Lyu, Patrick Heyer, Bart Goossens, Wilfried Philips
2022 Sensors  
Intermediate domain images are generated by translating the source images to mimic the target ones, acting as a better starting point for the parameter update of the pedestrian detector.  ...  on pseudo training labels.  ...  Effects of Illumination-Aware Label Fusion Pseudo labels provide functionality for training/fine-tuning a detector on the target domain without any manual training labels.  ... 
doi:10.3390/s22124416 pmid:35746199 pmcid:PMC9228565 fatcat:qdbvn5bmp5ctfaim2ta43h52fe

Let There be Light: Improved Traffic Surveillance via Detail Preserving Night-to-Day Transfer [article]

Lan Fu, Hongkai Yu, Felix Juefei-Xu, Jinlong Li, Qing Guo, Song Wang
2021 arXiv   pre-print
We propose to utilize style translation based StyleMix method to acquire pairs of day time image and nighttime image as training data for following nighttime to daytime image translation.  ...  To alleviate the detail corruptions caused by Generative Adversarial Networks (GANs), we propose to utilize Kernel Prediction Network (KPN) based method to refine the nighttime to daytime image translation  ...  This kind of image translation considers this problem as domain adaptation for model fine-tuning on synthetic nighttime images without labeling the nighttime data.  ... 
arXiv:2105.05011v1 fatcat:bfpju3ugavf3bbjv2ggfrawth4

Seeing Objects in dark with Continual Contrastive Learning [article]

Ujjal Kr Dutta
2022 arXiv   pre-print
As manually obtaining such a large labeled dataset may be infeasible, we suggest using synthetic images, to mimic different training image domains.  ...  Unfortunately, to build a well-performing detector across varying imaging conditions, one would require labeled training images (often in large numbers) from a plethora of corner cases.  ...  Also, the images are a good mix of both high-vs low-resolution, and near vs far away aspects of objects.  ... 
arXiv:2112.02891v3 fatcat:us5dxswl4bflnhwmv4pb6bmypm

Nighttime Data Augmentation using GAN for Improving Blind-Spot Detection

Hongjun Lee, Moonsoo Ra, Whoi-Yul Kim
2020 IEEE Access  
INDEX TERMS Data augmentation, domain adaptation, generative adversarial networks, blind-spot detection.  ...  Experiments on a real nighttime dataset demonstrate that the proposed framework improved the detection performance considerably in comparison with using daytime images only.  ...  [31] proposed pix2pixHD, which can generate high-resolution synthetic images. However, the paired images are almost impossible to obtain.  ... 
doi:10.1109/access.2020.2979239 fatcat:drla75x6mnbopikzqfbzhlhnxy

GPS-GLASS: Learning Nighttime Semantic Segmentation Using Daytime Video and GPS data [article]

Hongjae Lee, Changwoo Han, Seung-Won Jung
2022 arXiv   pre-print
Given GPS-aligned pairs of daytime and nighttime images, we perform cross-domain correspondence matching to obtain pixel-level pseudo supervision.  ...  Nighttime semantic segmentation is especially challenging due to a lack of annotated nighttime images and a large domain gap from daytime images with sufficient annotation.  ...  The spatial resolution of the original images and annotations is 2,048 × 1,024. We used the training dataset of 2,975 images as the labeled source domain S in the GPS-GLASS training stage.  ... 
arXiv:2207.13297v2 fatcat:artsb7g7vbe2fnivs7xkkll3mq

Map-Guided Curriculum Domain Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation [article]

Christos Sakaridis, Dengxin Dai, Luc Van Gool
2021 arXiv   pre-print
We address the problem of semantic nighttime image segmentation and improve the state-of-the-art, by adapting daytime models to nighttime without using nighttime annotations.  ...  images from a reference map and dark images to guide the label inference in the dark domains; 2) a novel uncertainty-aware annotation and evaluation framework and metric for semantic segmentation, including  ...  At the method level, this work adapts semantic segmentation models from daytime to nighttime, without annotations in the latter domain.  ... 
arXiv:2005.14553v2 fatcat:nzs4xiruuzbvtoopwkitccxi34

Semantic Segmentation with Low light Images by Modified CycleGAN-based Image Enhancement

Se Woon Cho, Na Rae Baek, Ja Hyung Koo, Muhammad Arsalan, Kang Ryoung Park
2020 IEEE Access  
The existing state-of-the-art segmentation methods show high performance for bright and clear images.  ...  In this study, we used low light databases generated by two road scene open databases that provide segmentation labels, which are the Cambridge-driving labeled video database (CamVid) and Karlsruhe Institute  ...  One is the original high-resolution image, and the others are mediumand low-resolution images downsampled by factors of 2 and 4, respectively.  ... 
doi:10.1109/access.2020.2994969 fatcat:g7zmakylljeabhkyr2ngx5htau

Night-to-Day Image Translation for Retrieval-based Localization [article]

Asha Anoosheh, Torsten Sattler, Radu Timofte, Marc Pollefeys, Luc Van Gool
2019 arXiv   pre-print
A recent class of neural models allows for realistic translation of images among visual domains with relatively little training data and, most importantly, without ground-truth pairings.  ...  We propose ToDayGAN - a modified image-translation model to alter nighttime driving images to a more useful daytime representation.  ...  These works show that being able to attain high-quality representations of images in the appearance of other domains is useful for tasks containing a shift in the data domain; irrelevant source-domain-specific  ... 
arXiv:1809.09767v2 fatcat:sumn7ggah5fvhkywayn55m3zyi

Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation [article]

Christos Sakaridis, Dengxin Dai, Luc Van Gool
2019 arXiv   pre-print
We instead address the problem of semantic segmentation of nighttime images and improve the state-of-the-art, by adapting daytime models to nighttime without using nighttime annotations.  ...  Most progress in semantic segmentation reports on daytime images taken under favorable illumination conditions.  ...  At the method level, this work instead adapts semantic segmentation models trained on daytime to nighttime without annotations in the latter domain.  ... 
arXiv:1901.05946v2 fatcat:qxmhkeyefjfkzbtyfq2bwvwrka

Time-of-Day Neural Style Transfer for Architectural Photographs [article]

Yingshu Chen, Tuan-Anh Vu, Ka-Chun Shum, Binh-Son Hua, Sai-Kit Yeung
2022 arXiv   pre-print
Inspired by recent successes in image-to-image translation methods, we aim to perform style transfer for architectural photographs.  ...  Our method comprises a segmentation module, a learning-based image-to-image translation module, and an image blending optimization module.  ...  We trained an InceptionV3 [59] classifier using our dataset with three target domain labels (i.e., golden, blue, and night).  ... 
arXiv:2209.05800v1 fatcat:5umcjcodwnbmnhpjm7rfgts2hy

HeatNet: Bridging the Day-Night Domain Gap in Semantic Segmentation with Thermal Images [article]

Johan Vertens, Jannik Zürn, Wolfram Burgard
2020 arXiv   pre-print
domain.  ...  We further employ a domain adaptation method to align the learned feature spaces across the domains and propose a novel two-stage training scheme.  ...  Transfer learning and domain adaptation approaches aim at narrowing the domain gap between a source domain, where supervised learning from labelled data is possible, to a target domain, where labelled  ... 
arXiv:2003.04645v1 fatcat:6l2peglozrdbbh3m53sknplnia

ROMA: Cross-Domain Region Similarity Matching for Unpaired Nighttime Infrared to Daytime Visible Video Translation [article]

Zhenjie Yu, Kai Chen, Shuang Li, Bingfeng Han, Chi Harold Liu, Shuigen Wang
2022 arXiv   pre-print
Although, the domain gaps between unpaired nighttime infrared and daytime visible videos are even huger than paired ones that captured at the same time, establishing an effective translation mapping will  ...  To be specific, ROMA could efficiently translate the unpaired nighttime infrared videos into fine-grained daytime visible ones, meanwhile maintain the spatiotemporal consistency via matching the cross-domain  ...  Daytime Visible Videos. Since we want high-quality translated daytime visible results, we shoot the daytime visible videos on a clear day.  ... 
arXiv:2204.12367v1 fatcat:cqej242nuvai3f4mrtanmlaysa
« Previous Showing results 1 — 15 out of 4,091 results