Article
Optics
Jiangxin Yang, Binjie Ding, Zewei He, Gang Pan, Yanpeng Cao, Yanlong Cao, Qian Zheng
Summary: This article introduces a new reflectance decomposition method called ReDDLE-Net, which can simultaneously consider the contributions of diffuse and specular reflectance for light estimation and achieves better performance compared to other methods.
Article
Computer Science, Hardware & Architecture
Xin Jin, Xingfan Zhu, Xinxin Li, Kejun Zhang, Xiaodong Li, Xiaokun Zhang, Quan Zhou, Shujiang Xie, Xi Fang
Summary: In this paper, the fourth-order spherical harmonic function is used to model illumination in indoor scenes, with 48 spherical harmonic coefficients representing the entire scene. High dynamic range environment maps are adopted to enhance illumination quality, with the goal of predicting spherical harmonic coefficients of corresponding high dynamic range images using low dynamic range images. Experimental results demonstrate the effectiveness of the method in accurately predicting spherical harmonic coefficients and producing realistic luminance.
COMPUTERS & ELECTRICAL ENGINEERING
(2021)
Article
Computer Science, Artificial Intelligence
Hao Zhang, Jiayi Ma
Summary: This paper proposes a multi-exposure fusion network for the unsupervised generation of high dynamic range-like images. A new intrinsic image decomposition (IID) network is developed to produce different components, including reflectance, shading, and color, from source images. Three fusion sub-networks are designed to process these components and generate pleasing images with rich structures, reasonable lighting, and suitable color.
INFORMATION FUSION
(2023)
Article
Computer Science, Artificial Intelligence
Junqing Huang, Michael Ruzhansky, Qianying Zhang, Haihui Wang
Summary: This article presents a novel intrinsic image transfer (IIT) algorithm for image illumination manipulation, which creates a local image translation between two illumination surfaces. The model is constructed based on an optimization framework and utilizes illumination, reflectance, and content losses. These losses are defined on sub-layers factorized by an intrinsic image decomposition and reduced under the illumination-invariant reflectance prior knowledge. The article also demonstrates the algorithm's versatility in illumination compensation, image enhancement, tone mapping, and HDR image compression, showcasing high-quality results on natural image datasets.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Chemistry, Analytical
Omar Elezabi, Sebastien Guesney-Bodet, Jean-Baptiste Thomas
Summary: Spectral Filter Array cameras provide a fast and portable solution for spectral imaging. This study investigates texture classification methods applied directly to the raw image, and shows that the Convolutional Neural Network outperforms other methods in terms of classification performance. The study also demonstrates the model's ability to adapt and scale for different environmental conditions.
Article
Geochemistry & Geophysics
Wen Xie, Yanfeng Gu, Tianzhu Liu
Summary: This article proposes a two-stream encoder-decoder network for the single hyperspectral (HS) task, which includes a reflectance estimation subnetwork (RES) and a shading estimation subnetwork (SES). The network introduces three physical losses to enhance performance and ensures the estimated intrinsic components are physically correct. Experimental results demonstrate that the proposed network outperforms current available learning or optimization-based approaches.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Mahdi Yousefan, Hamid Esmaeili Najafabadi, Hossein Amirkhani, Henry Leung, Vahid Hajihashemi
Summary: This paper proposes a novel framework for transferred deep learning-based anomaly detection in hyperspectral images, which includes PCA dimension reduction, CNN training, and OAF algorithm utilization. The framework outperforms many state-of-the-art methods and demonstrates excellent domain adaptability.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Software Engineering
Salma Jiddi, Philippe Robert, Eric Marchand
Summary: This article introduces a novel framework to recover the position and color of multiple light sources as well as the specular reflectance of real scene surfaces using a single RGB-D camera. The approach detects and incorporates information from both specular reflections and cast shadows to achieve its goal.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2022)
Article
Biology
Muhammad Asif, Hong Song, Lei Chen, Jian Yang, Alejandro F. Frangi
Summary: An automatic specular reflection detection method based on intrinsic image layer separation (IILS) is proposed in this study, which outperforms other state-of-the-art methods in terms of detection performance. The evaluation results demonstrate better qualitative and quantitative assessments compared with other methods, showing promise as a preprocessing step for further analysis of endoscopic images.
COMPUTERS IN BIOLOGY AND MEDICINE
(2021)
Article
Computer Science, Information Systems
Saleh Javadi, Mattias Dahl, Mats I. Pettersson
Summary: This article investigates the use of 3D feature maps to enhance the performance of deep neural networks in vehicle detection in aerial images. By evaluating different base networks and their combination with YOLOv3, as well as generating 3D depth maps and training a fully connected neural network for vehicle detection.
Article
Computer Science, Software Engineering
Zhengfei Kuang, Kyle Olszewski, Menglei Chai, Zeng Huang, Panos Achlioptas, Sergey Tulyakov
Summary: This study presents a novel method to acquire object representations from online image collections, capturing high-quality geometry and material properties. By using a multi-stage approach, the surface geometry and camera parameters are inferred, and the training efficiency is improved by leveraging coarse foreground object masks. A robust normal estimation technique is employed to eliminate geometric noise while retaining crucial details. The resulting object acquisition framework is highly modular and efficient, and is advantageous in capturing high-quality geometry and appearance properties useful for rendering applications.
ACM TRANSACTIONS ON GRAPHICS
(2022)
Article
Computer Science, Artificial Intelligence
Qian Zhang, Shuang Lu, Lei Liu, Yi Liu, Jing Zhang, Daoyuan Shi
Summary: This paper focuses on enhancing the color of garden landscape images in unfavorable shooting environments through modified dynamic threshold and convolutional neural network, achieving color restoration and clarity enhancement of low illumination GLIs.
TRAITEMENT DU SIGNAL
(2021)
Article
Multidisciplinary Sciences
Mehwish Iqbal, Muhammad Mohsin Riaz, Abdul Ghafoor, Attiq Ahmad
Summary: An illumination normalization algorithm for face images is proposed, which involves enhancing low light, normalizing the reflection layer, removing noise, and enhancing details to improve image quality. The algorithm corrects contrast by histogram truncation and extracts shape from shading through a guided filter to achieve illumination normalization of a face image. Simulation results on multiple face datasets demonstrate the effectiveness of this approach.
ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING
(2021)
Article
Computer Science, Artificial Intelligence
Yongjie Zhu, Chen Li, Si Li, Boxin Shi, Yu-Wing Tai
Summary: In this paper, we propose HyFRIS-Net to jointly estimate the hybrid reflectance and illumination models, as well as the refined face shape in a pre-defined texture space from a single unconstrained face image. Our method ensures efficient learning of photometric face appearance modeling in both parametric and non-parametric spaces by enforcing reflectance consistency and face identity constraints. We recover an occlusion-free face albedo with disambiguated color from the illumination color. The network is trained in a self-evolving manner to achieve general applicability on real-world data. Comprehensive evaluations demonstrate the advantages of HyFRIS-Net in modeling photo-realistic face albedo, illumination, and shape.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Geography
Jialin Li, Ningchuan Xiao
Summary: This article explores the use of artificial intelligence and machine learning methods in understanding maps, specifically in tasks such as map identification, geographic region recognition, and projection recognition. The results show that pretrained CNN models achieve the highest performance, with an accuracy rate above 90%. However, the performance may decline when tested with systematically distorted maps, indicating both promises and limitations of the current machine learning approaches to cartographic recognition.
ANNALS OF THE AMERICAN ASSOCIATION OF GEOGRAPHERS
(2023)
Article
Computer Science, Artificial Intelligence
Yu Liu, Tinne Tuytelaars
Summary: Discovering novel visual categories from unlabeled images is crucial for intelligent vision systems, and we propose a residual-tuning approach to overcome the tradeoff between preserving features on labeled data and adapting features on unlabeled data. Our method achieves consistent and considerable gains on benchmark tests, reducing the performance gap to fully supervised learning setup.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Agriculture, Multidisciplinary
Tim Van De Looverbosch, Jiaqi He, Astrid Tempelaere, Klaas Kelchtermans, Pieter Verboven, Tinne Tuytelaars, Jan Sijbers, Bart Nicolai
Summary: X-ray radiography has been investigated as a technique for internal quality inspection of pears in storage, with multiple deep anomaly detection methods showing effectiveness in detecting pears with internal cavity and browning disorders. The best performing methods were found to be on par with a state-of-the-art multisensor disorder detection method.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2022)
Article
Medicine, General & Internal
Riaan Zoetmulder, Agnetha A. E. Bruggeman, Ivana Isgum, Efstratios Gavves, Charles B. L. M. Majoie, Ludo F. M. Beenen, Diederik W. J. Dippel, Nikkie Boodt, Sanne J. den Hartog, Pieter J. van Doormaal, Sandra A. P. Cornelissen, Yvo B. W. E. M. Roos, Josje Brouwer, Wouter J. Schonewille, Anne F. V. Pirson, Wim H. van Zwam, Christiaan van der Leij, Rutger J. B. Brans, Adriaan C. G. M. van Es, Henk A. Marquering
Summary: In this study, an automatic method for thrombus localization and segmentation on CT images in patients with posterior circulation stroke (PCS) was developed. The method achieved good results in localizing and segmenting thrombi. Restricting the volume-of-interest (VOI) to the brainstem improved the precision and recall of thrombus localization.
Article
Computer Science, Artificial Intelligence
Eli Verwimp, Kuo Yang, Sarah Parisot, Lanqing Hong, Steven McDonagh, Eduardo Perez-Pellitero, Matthias De Lange, Tinne Tuytelaars
Summary: In this paper, a new Continual Learning benchmark for Autonomous Driving (CLAD) is introduced, focusing on object classification and object detection problems. The benchmark utilizes SODA10M, a large-scale dataset related to autonomous driving. Existing continual learning benchmarks are reviewed and discussed, showing that most of them are extreme cases. Online classification benchmark CLAD-C and domain incremental continual object detection benchmark CLAD-D are introduced. The inherent difficulties and challenges are examined through a survey of top-3 participants in a CLAD-challenge workshop at ICCV 2021. Possible pathways to improve the current state of continual learning and promising directions for future research are discussed.
Article
Agronomy
Astrid Tempelaere, Tim Van De Looverbosch, Klaas Kelchtermans, Pieter Verboven, Tinne Tuytelaars, Bart Nicolai
Summary: This study proposes a method to generate synthetic CT images using a conditional cGAN to overcome the challenges of obtaining large annotated datasets. The performance of the predictor was evaluated quantitatively and visually, showing that the cGAN effectively generated CT images of healthy and defective fruit based on annotations.
POSTHARVEST BIOLOGY AND TECHNOLOGY
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Thomas Verelst, Paul K. Rubenstein, Marcin Eichner, Tinne Tuytelaars, Maxim Berman
Summary: Multi-label image classification is more practical for real-world scenarios than single-label classification due to the presence of multiple objects in natural images. However, annotating every object of interest is time-consuming and expensive. In this study, we propose an Expected Negative loss to train multi-label classifiers using datasets where each image is annotated with a single positive label. To handle the uncertainty of other classes, we generate a set of expected negative labels based on prediction consistency. Additionally, we introduce a novel spatial consistency loss to improve supervision by maintaining consistent spatial feature maps for each training image. Our experiments on various datasets demonstrate the effectiveness of the Expected Negative loss in combination with consistency and spatial consistency losses, and we achieve improved multi-label classification mAP on ImageNet-1K using the ReaL multi-label validation set.
2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV)
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Abhishek Jha, Soroush Seifi, Tinne Tuytelaars
Summary: In active visual exploration, it is crucial to sample informative local observations for modeling global context. This paper proposes the use of vision transformers instead of CNNs for such agents and introduces a transformer-based active visual sampling model called SimGlim. The model utilizes the transformer's self-attention architecture to predict the best next location based on the current observable environment. Experimental results demonstrate the effectiveness of the proposed method in image reconstruction and comparisons against existing methods are provided. Ablation studies are also conducted to analyze the importance of design choices in the overall architecture.
2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV)
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Abhishek Jha, Badri Patro, Luc Van Gool, Tinne Tuytelaars
Summary: This paper proposes a novel regularization method called COB to improve the information content of the joint space in visual question answering models. It reduces redundancy by minimizing the correlation between learned feature components, disentangling semantic concepts. The model aligns the joint space with the answer embedding space and shows improved accuracy on VQA datasets.
2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV)
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Tim Lebailly, Tinne Tuytelaars
Summary: The downstream accuracy of self-supervised methods depends on the proxy task and the quality of gradients extracted during training. Incorporating local cues in the proxy task can improve model accuracy on downstream tasks. We propose a geometric approach for matching local representations in self-distillation, which outperforms similarity-based methods, especially in low-data regimes. However, similarity-based matchings are highly detrimental to model performance in low-data regimes compared to the baseline without local self-distillation.
2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV)
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Tingyu Qu, Tinne Tuytelaars, Marie-Francine Moens
Summary: This paper revisits the weakly supervised cross-modal face-name alignment task and proposes SECLA and SECLA-B models. These models use appropriate loss functions to learn the alignments between names and faces in a neural network setting. SECLA maximizes the similarity scores between faces and names in a weakly supervised fashion, while SECLA-B learns to align names and faces from easy to hard cases, further improving the performance.
2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV)
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Thomas Stegmuller, Tim Lebailly, Behzad Bozorgtabar, Tinne Tuytelaars, Jean-Philippe Thiran
Summary: In this paper, we propose a method for learning dense visual representations without labels by discovering and segmenting the semantics of views through an online clustering mechanism. The resulting method is highly generalizable and does not require cumbersome pre-processing steps.
2023 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR
(2023)
Article
Computer Science, Artificial Intelligence
Matthias De Lange, Rahaf Aljundi, Marc Masana, Sarah Parisot, Xu Jia, Ales Leonardis, Greg Slabaugh, Tinne Tuytelaars
Summary: This article introduces the application of artificial neural networks in continual learning, focusing on task incremental classification. It proposes a new framework for continually evaluating the stability-plasticity trade-off of the network and performs experimental comparisons of 11 state-of-the-art continual learning methods, evaluating their strengths and weaknesses by considering different benchmark datasets.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Proceedings Paper
Automation & Control Systems
Klaas Kelchtermans, Tinne Tuytelaars
Summary: The gap between simulation and the real world hampers the application of machine learning in computer vision and reinforcement learning. This study addresses this issue by focusing on camera-based navigation and utilizing various techniques to successfully bridge the gap between simulation and the real world.
2022 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Shipeng Yan, Lanqing Hong, Hang Xu, Jianhua Han, Tinne Tuytelaars, Zhenguo Li, Xuming He
Summary: This study focuses on learning VLP models with sequential chunks of image-text pair data and proposes pseudo text replay and multi-modal knowledge distillation to tackle the forgetting issue in continual learning. The experiments demonstrate the superiority of the proposed method in zero-shot image classification and image-text retrieval tasks.
COMPUTER VISION, ECCV 2022, PT XXXVI
(2022)
Article
Computer Science, Artificial Intelligence
Gido M. van de Ven, Tinne Tuytelaars, Andreas S. Tolias
Summary: Deep neural networks face challenges in continual learning, with different scenarios of continual learning having varying challenges and effectiveness. Distinguishing between task-incremental, domain-incremental, and class-incremental learning is an important foundation for organizing the continual learning field.
NATURE MACHINE INTELLIGENCE
(2022)