Article
Computer Science, Information Systems
Zhen Qin, Qingya Chen, Yi Ding, Tianming Zhuang, Zhiguang Qin, Kim-Kwang Raymond Choo
Summary: This paper proposes a novel approach called ObjectVariedGAN to handle geometric translation in image-to-image transformation. The approach focuses on maintaining the shape of foreground objects and utilizes feature similarity loss and cycle-consistency loss to generate the desired output without requiring paired training data.
INFORMATION PROCESSING & MANAGEMENT
(2022)
Article
Environmental Sciences
Joanna Zawadzka, Ian Truckell, Abdou Khouakhi, Monica Rivas Casado
Summary: The study utilized UAV-captured high resolution imagery for rapid assessment of flood damage, automatically detecting debris associated with residential housing damages through image segmentation and classifiers, achieving a good level of accuracy.
Article
Environmental Sciences
Yunfei Han, Ping Wang, Yongguo Zheng, Muhammad Yasir, Chunmei Xu, Shah Nazir, Md Sakaouth Hossain, Saleem Ullah, Sulaiman Khan
Summary: This study utilizes object-oriented classification to extract landslide data from high-resolution remote sensing data and explores the impact of geology, lithology, rainfall, and human activities on landslide occurrence. The study found a Kappa coefficient of 0.76, landslide extraction accuracy of 79.8%, and an overall classification accuracy of 87%. The causes of landslides are discussed and early warning information for unknown landslides can be obtained through feature analysis.
Article
Computer Science, Artificial Intelligence
Guangtao Nie, Hua Huang
Summary: This article proposes a method to solve the discontinuity problem in multi-oriented object detection by encoding the object with double horizontal rectangles (DHRec). By arranging the coordinates of the four vertices in left-right and top-down order, the uniqueness of the encoding is ensured. The method uses area ratios to guide the decoding of the object. Experimental results show that the proposed method can accurately detect objects of arbitrary orientation and outperforms existing methods.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Environmental Sciences
Chongyang Wei, Weiping Ni, Yao Qin, Junzheng Wu, Han Zhang, Qiang Liu, Kenan Cheng, Hui Bian
Summary: Compared with general object detection in natural images, oriented object detection in remote sensing images is challenging due to arbitrary orientations. Existing CNN-based methods often adopt complex and inefficient approaches to model variant orientations. In this paper, we propose a lightweight approach to generate oriented proposals and extract rotation-invariant features. Our method achieves state-of-the-art accuracy while reducing the model size by 40%.
Article
Chemistry, Analytical
Tai-Hung Lin, Chih-Wen Su
Summary: This paper proposes a simple and efficient oriented object detector based on the YOLOv4 architecture. By regressing the offset of an object's front point instead of its angle or corners, the discontinuous boundary problem caused by angular periodicity or corner order is avoided. The introduction of the intersection over union (IoU) correction factor makes the training process more stable. Experimental results show that the proposed method outperforms other methods in terms of detection speed and accuracy.
Article
Environmental Sciences
Linhai Wei, Chen Zheng, Yijun Hu
Summary: In this paper, a new two-stage object detector called Faster R-CNN-NeXt with RoI-Transformer is proposed for detecting small and densely packed objects with complicated orientations in remote sensing aerial images. By introducing a scaled smooth L1 loss function, the proposed detector improves the detection performance and reduces the impact of detection error variance. Experimental results show that the proposed detector outperforms other methods on popular datasets.
Article
Physics, Multidisciplinary
Junfeng Lv, Tian Hui, Yongfeng Zhi, Yuelei Xu
Summary: With the development of image technology, automatic image captioning for infrared images has become increasingly important in various industries. This task is widely used in night security and understanding night scenes. However, generating captions for infrared images remains challenging due to differences in image features and the complexity of semantic information. To improve the correlation between descriptions and objects, we introduced YOLOv6 and LSTM as encoder-decoder structure and proposed an object-oriented attention method for infrared image captioning.
Article
Environmental Sciences
Zhifeng Xiao, Kai Wang, Qiao Wan, Xiaowei Tan, Chuan Xu, Fanfan Xia
Summary: The study introduces a detection method that adapts anchoring based on sample balance, selecting candidate anchors through horizontal IoU and using an adaptive threshold module to maintain a balance between positive and negative anchors, ultimately achieving better performance on a public aerial image dataset.
Review
Oncology
Peter Bankhead
Summary: The potential of quantitative image analysis and artificial intelligence in digital pathology is highlighted. However, the lack of available software and the complexity of existing methods hinder widespread adoption. This review emphasizes the need for collaboration and multidisciplinary approaches in developing new algorithms and calls for greater attention to openness, implementation, and usability. The interaction between digital pathology and the bioimage analysis community is seen as beneficial in terms of data sharing and idea exchange.
JOURNAL OF PATHOLOGY
(2022)
Article
Geosciences, Multidisciplinary
Fangjian Liu, Lei Dong, Xueli Chang, Xinyi Guo
Summary: Remote sensing image classification plays a crucial role in urban development and planning. This research proposes an object-oriented convolutional neural network (OCNN) method and compares it with SVM and convolutional neural network. The experimental results demonstrate that the OCNN method outperforms the others in terms of classification accuracy and image continuity.
FRONTIERS IN EARTH SCIENCE
(2022)
Article
Environmental Sciences
Qi Ming, Lingjuan Miao, Zhiqiang Zhou, Junjie Song, Xue Yang
Summary: This paper proposes a sparse label assignment strategy to improve sample assignment in object detection, utilizes a position-sensitive feature pyramid network with a coordinate attention module to extract position-sensitive features, and introduces a distance rotated IoU loss to enhance bounding box regression.
Article
Engineering, Electrical & Electronic
Yudong Wang, Jichang Guo, Ruining Wang, Wanru He, Chongyi Li
Summary: This paper proposes a task-oriented image enhancement network (TIENet) to directly improve the performance of degraded object detection by enhancing the degraded images. TIENet is a zero-reference enhancement network that adds a detection-favorable structure image to the original degraded image. The experiments and evaluations demonstrate that the proposed framework achieves significant improvements on classic detectors.
SIGNAL IMAGE AND VIDEO PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Yongchao Xu, Mingtao Fu, Qimeng Wang, Yukang Wang, Kai Chen, Gui-Song Xia, Xiang Bai
Summary: A framework for detecting multi-oriented objects is proposed in this paper, which accurately describes multi-oriented objects by sliding bounding box vertices and introducing additional target variables, achieving superior performances on multiple object detection benchmarks.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Environmental Sciences
Yangyang Li, Heting Mao, Ruijiao Liu, Xuan Pei, Licheng Jiao, Ronghua Shang
Summary: A lightweight keypoint-based oriented object detector for remote sensing images is proposed in this paper, which improves detection performance by introducing a semantic transfer block and an adaptive Gaussian kernel, and obtains a lightweight student network using distillation loss associated with object detection. Experimental results show that the method adapts to objects of different scales, obtains accurate bounding boxes, and reduces the influence of complex backgrounds. The comparison with mainstream methods demonstrates comparable performance under lightweight conditions.