This research investigates the application of iterative point-based sparse annotations for semantic segmentation of cars in remote-sensing imagery, to mitigate the challenges associated with laborious and expensive data labeling processes. Car labeling considered the selection of point shapefiles in the Geographic Information System with a value of 1 for the specific class of car, 0 to background and a value of -1 outside the intended target. The semantic segmentation model was the U-Net architecture, with an Efficient-net-B7 backbone and a modified cross-entropy loss function. The experimental evaluation uses the BSB Vehicle Dataset, encompassing two classes (background and vehicles). The results showcase promising improvements, particularly in error-prone classes, as more samples are iteratively added during training. This approach presents a viable and time-efficient alternative for dataset creation, leveraging sparse annotations that are incrementally enhanced. Our pipeline included five iterative rounds in which the IoU increased more than 30% from the first round to the fifth, achieving 60% IoU with 0.059% of the total annotated data. Showing a very good performance with less than 0.1% of pixels annotated. This research advances the field by proposing a rapid and cost-effective method for generating high-quality datasets in remote sensing.

This study addresses the challenge of amodal segmentation in computer vision, a change in basic assumptions towards perceiving objects holistically, even when partially occluded, deviating from the traditional modal perspective that predominantly focuses on visible elements. Thus, we propose a new approach for the amodal segmentation of top-view aerial images, with particular attention to the first layer of elements, constituted by asphalt and natural soils, normally occluded by different objects (trees, buildings, and vehicles). This proposed methodology is data-centric, assigning weights to specific image sections and distinguishing non-visible elements. The best model used the U-Net architecture with Efficient-net-B7 as the backbone and can accurately classify occluded segments, achieving an Intersection over Union (IoU) greater than 80% for most classes. The developed method provides a basis for exploring amodal segmentation based on data-centric models, impacting our understanding of complex and occlusion-prone environments, such as urban environments.

This study investigates the application of iterative sparse annotations for semantic segmentation in remote-sensing imagery, focusing on minimizing the laborious and expensive data labeling process. By leveraging Geographic Information Systems (GIS), we implemented circular polygon shapefiles to label portions of each class, attributing a value of -1 outside these polygons. The model training used the simplified BSB Aerial Dataset with eight classes. The semantic segmentation model was U-Net architecture with the Efficient-net-B7 backbone and a modified cross-entropy loss function. Our results showed promising improvement, particularly in error-prone classes, with the iterative addition of more samples. This approach suggests a quicker method for dataset creation using sparse, iteratively enhanced annotations. Future work will aim to implement further iterative rounds to approximate the results of continuous labeling, thereby enhancing the efficiency of semantic segmentation in large-scale remote-sensing images.

This research aims to investigate the use of semantic segmentation and Shuttle Radar Topography Mission (SRTM) data in detecting natural karst depressions developed on the carbonate rocks of the Neoproterozoic Bambuí Group in Western Bahia, Brazil. The study area is a karst landscape containing depressions enclosed in limestone, many forming lakes. The methodology had the following steps: (a) visual interpretation of karst depressions from Sentinel-2 and OLI-Landsat 8 images; (b) generation of DEM-based sink depth plus nine morphometric attributes; (c) selection of 128x128-pixel samples for training (1600), validation (400), and testing (400) considering two channels (DEM and sink depth based on DEM) and eleven channels (the two previous ones and the morphometric attributes); and (d) semantic segmentation using U-Net architecture with EfficientNet-B7 backbone. The accuracy metrics were 98.26, 72.82, 79.50, 79.16, and 65.51 for OA, precision, recall, F-score, and IoU when considering SRTM plus morphometric attributes (11 channels).

Panoptic segmentation combines instance and semantic seg-mentation, enabling the classification of objects and back-grounds. It is still a method little explored in the remote sensing field, mostly due to the difficulty of generating the data. Moreover, the beach areas have great interest due to many objects and elements that may guide public policies. In this regard, we propose the first study on beach areas using panop-tic segmentation and the first panoptic segmentation study using multispectral data. We used the Gram-Schmidt pan-sharpening method for the multispectral bands and created a dataset with 850 samples with 128x 128 dimensions in the COCO panoptic annotation format. To evaluate the dataset, the Panoptic-FPN was used with modifications in the input (changing from three to eight channels). Results show 59.43 Panoptic Quality (PQ), 77.96 Segmentation Quality (SQ), and 75.07 Recognition Quality (RQ).

2D representations of 3D scenes generate occlusions among different targets. Understanding targets by only seeing parts of them is referred to as an amodal perception, which is still unexplored in remote sensing. Thus, we propose integrating this concept using peculiarities of remote sensing Nadir images to classify non-visible targets at a pixel level. Nadir images present a hierarchical order of occlusions, allowing us to separate different layers. We developed a dataset with 600 images and three classes (roads, vehicles, and trees) with in-dependent labelling for each class. Any semantic segmentation model is suitable for this task, but we explored the U-net architecture with three backbones (Efficient-net-B7, ResNet-101, and ResNeXt-101). The evaluation considered the IoU metric, providing 80% for the best model (Efficient-net-B7). Future studies aim to extend this approach by introducing competing classes among each layer and increasing the number of samples and categories.

This letter proposes a novel method to obtain panoptic predictions by extending the semantic segmentation task with a few non-learning image processing steps, presenting the following benefits: 1) annotations do not require a specific format [e.g., common objects in context (COCO)]; 2) fewer parameters (e.g., single loss function and no need for object detection parameters); and 3) a more straightforward sliding windows implementation for large image classification (still unexplored for panoptic segmentation). Semantic segmentation models do not individualize touching objects, as their predictions can merge; i.e., a single polygon represents many targets. Our method overcomes this problem by isolating the objects using borders on the polygons that may merge. The data preparation requires generating a one-pixel border, and for unique object identification, we create a list with the isolated polygons, attribute a different value to each one, and use the expanding border (EB) algorithm for those with borders. Although any semantic segmentation model applies, we used the U-Net with three backbones (EfficientNet-B5, EfficientNet-B3, and EfficientNet-B0). The results show that the following hold: 1) the EfficientNet-B5 had the best results with 70% mean intersection over union (mIoU); 2) the EB algorithm presented better results for better models; 3) the panoptic metrics show a high capability of identifying things and stuff with 65 panoptic quality (PQ); and 4) the sliding windows on a$2560\times 2560$-pixel area has shown promising results, in which the ratio of merged objects by correct predictions was lower than 1% for all classes.

Center pivots are a modern irrigation technique mainly applied in precision agriculture, once it has high efficiency in water consumption and low labor workers when compared to traditional irrigation methods. Knowing their location is valuable since monitoring, evaluating, and estimating essential features in the lands becomes easier, remote sensing is a robust tool to act upon this kind of problem. To identify center pivots, we used a deep residual U-Net with a pixel comparison at image reconstruction to enhance results. We obtained a validation loss of 0.19, which adds up with pixel comparison. Results were satisfactory, with 2070 correct identifications from a total of 2109 center pivots (98.15%). Future studies to improve these results would require more data in different places and seasons.

Time series of satellite images allows better monitoring and detecting the dynamics of urban growth. The objective of this research is to detect the urban area between the city of Rio de Janeiro and São Paulo in the period of 2014-2015 using MODIS digital time series. We used the product MOD13Q1 referring to the Normalized Difference Vegetation Index (NDVI) 16-day composite data, with spatial resolution of 250 meters. The high variety of elements imposes a great difficulty in mapping urban areas. Therefore, we calculated the nearest neighbor using the Euclidian distance for each time signature. The different image group of the urban targets were into a single image considering the minimum value of each pixel within the set. Therefore, a limit value separated the urban areas from the rest. This methodology allowed the detection of urban areas considering their diversity. The algorithm is written in C ++ language.