Unplanned urban settlements exist worldwide. The geospatial information of these areas is critical for urban management and reconstruction planning but usually unavailable. Automatically characterizing individual buildings in the unplanned urban village using remote sensing imagery is very challenging due to complex landscapes and high-density settlements. The newly emerging deep learning method provides the potential to characterize individual buildings in a complex urban village. This study proposed an urban village mapping paradigm based on U-net deep learning architecture. The study area is located in Guangzhou City, China. The Worldview satellite image with eight pan-sharpened bands at a 0.5-m spatial resolution and building boundary vector file were used as research purposes. There are ten sites of the urban villages included in this scene of the Worldview image. The deep neural network model was trained and tested based on the selected six and four sites of the urban village, respectively. Models for building segmentation and classification were both trained and tested. The results indicated that the U-net model reached overall accuracy over 86% for building segmentation and over 83% for the classification. The F₁-score ranged from 0.9 to 0.98 for the segmentation, and from 0.63 to 0.88 for the classification. The Interaction over Union reached over 90% for the segmentation and 86% for the classification. The superiority of the deep learning method has been demonstrated through comparison with Random Forest and object-based image analysis. This study fully showed the feasibility, efficiency, and potential of the deep learning in delineating individual buildings in the high-density urban village. More importantly, this study implied that through deep learning methods, mapping unplanned urban settlements could further characterize individual buildings with considerable accuracy.

Single-class object detection, which focuses on identifying, counting, and tracking a specific animal species, plays a vital role in optimizing farm operations. However, dense occlusion among individuals in group activity scenarios remains a major challenge. To address this, we propose YOLO-SDD, a dense detection network designed for single-class densely populated scenarios. First, we introduce a Wavelet-Enhanced Convolution (WEConv) to improve feature extraction under dense occlusion. Following this, we propose an occlusion perception attention mechanism (OPAM), which further enhances the model’s ability to recognize occluded targets by simultaneously leveraging low-level detailed features and high-level semantic features, helping the model better handle occlusion scenarios. Lastly, a Lightweight Shared Head (LS Head) is incorporated and specifically optimized for single-class dense detection tasks, enhancing efficiency while maintaining high detection accuracy. Experimental results on the ChickenFlow dataset, which we developed specifically for broiler detection, show that the n, s, and m variants of YOLO-SDD achieve AP_50:95 improvements of 2.18%, 2.13%, and 1.62% over YOLOv8n, YOLOv8s, and YOLOv8m, respectively. In addition, our model surpasses the detection performance of the latest real-time detector, YOLOv11. YOLO-SDD also achieves state-of-the-art performance on the publicly available GooseDetect and SheepCounter datasets, confirming its superior detection capability in crowded livestock settings. YOLO-SDD’s high efficiency enables automated livestock tracking and counting in dense conditions, providing a robust solution for precision livestock farming.

Today, large-scale Penaeus monodon farms no longer incubate eggs but instead purchase larvae from large-scale hatcheries for rearing. The accurate counting of tens of thousands of larvae in these transactions is a challenging task due to the small size of the larvae and the highly congested scenes. To address this issue, we present the Penaeus Larvae Counting Strategy (PLCS), a simple and efficient method for counting Penaeus monodon larvae that only requires a smartphone to capture images without the need for any additional equipment. Our approach treats two different types of keypoints as equip keypoints based on keypoint regression to determine the number of shrimp larvae in the image. We constructed a high-resolution image dataset named Penaeus_1k using images captured by five smartphones. This dataset contains 1420 images of Penaeus monodon larvae and includes general annotations for three keypoints, making it suitable for density map counting, keypoint regression, and other methods. The effectiveness of the proposed method was evaluated on a real Penaeus monodon larvae dataset. The average accuracy of 720 images with seven different density groups in the test dataset was 93.79%, outperforming the classical density map algorithm and demonstrating the efficacy of the PLCS.

Grassland is a widely distributed land use type that provides essential resources for livestock production and serves as a crucial ecosystem component. Over the past decades, grassland has significantly degraded and shrunk due to climate change and human activities, with ongoing changes in its area. This study utilized the Google Earth Engine (GEE) and the Res-UNet++ model to analyze the phenological and spectral characteristics of grasslands in Landsat images for long-term annual monitoring. Additionally, the LandTrendr algorithm was utilized to correct long-term time series data for grasslands, yielding a map of grassland distribution in Inner Mongolia from 1991 to 2020. The results indicate that the overall spatial accuracy from 1991 to 2020 exceeded 96%, with a Kappa coefficient over 0.92, demonstrating high monitoring accuracy. In particular, the 2019 grassland monitoring area showed high consistency with data from the Third National Land Survey (TNLS), with a coefficient of determination (R^2) reaching 0.97, reflecting the high accuracy and reliability.

Fish fry counting has been vital in fish farming, but current computer-based methods are not feasible enough to accurately and efficiently calculate large number of fry in a single count due to severe occlusion, dense distribution and the small size of fish fry. To address this problem, we propose the deconvolution enhancement keypoint network (DEKNet), a method for fish fry counting that features a single-keypoint approach. This novel approach models the fish fry as a point located in the central part of the fish head, laying the foundation for our innovative counting strategy. To be specific, first, a fish fry feature extractor (FFE) characterized by parallel dual branches is designed for high-resolution representation. Next, two identical deconvolution modules (TDMs) are added to the generation head for a high-quality and high-resolution keypoint heatmap with the same resolution size as the input image, thus facilitating the precise counting of fish fry. Then, the local peak value of the heatmap is obtained as the keypoint of the fish fry, so the number of these keypoints with coordinate information equals the number of fry, and the coordinates of the keypoint can be used to locate the fry. Finally, FishFry-2023, a large-scale fish fry dataset, is constructed to evaluate the effectiveness of the method proposed by us. Experimental results show that an accuracy rate of 98.59% was accomplished in fish fry counting. Furthermore, DEKNet achieved a high degree of accuracy on the Penaeus dataset (98.51%) and an MAE of 13.32 on a public dataset known as Adipocyte Cells. The research outcomes reveal that DEKNet has superior comprehensive performance in counting accuracy, the number of parameters and computational effort.

Affected by shooting angle and light intensity, shooting through transparent media may cause light reflections in an image and influence picture quality, which has a negative effect on the research of computer vision tasks. In this paper, we propose a Residual Attention Based Reflection Removal Network (RABRRN) to tackle the issue of single image reflection removal. We hold that reflection removal is essentially an image separation problem sensitive to both spatial and channel features. Therefore, we integrate spatial attention and channel attention into the model to enhance spatial and channel feature representation. For a more feasible solution to solve the problem of gradient disappearance in the iterative training of deep neural networks, the attention module is combined with a residual network to design a residual attention module so that the performance of reflection removal can be ameliorated. In addition, we establish a reflection image dataset named the SCAU Reflection Image Dataset (SCAU-RID), providing sufficient real training data. The experimental results show that the proposed method achieves a PSNR of 23.787 dB and an SSIM value of 0.885 from four benchmark datasets. Compared with the other most advanced methods, our method has only 18.524M parameters, but it obtains the best results from test datasets.