Dataset Information

Synthetic car dataset for vehicle detection: Integrating aerial and satellite imagery.

ABSTRACT: Vehicle detection is a very important aspect of computer vision application to aerial and satellite imagery, facilitating activities such as instance counting, velocity estimation, traffic predictions, etc. The feasibility of accurate vehicle detection often depends on limited training datasets, requiring a lot of manual work in collection and annotation tasks. Furthermore, there are no known publicly available datasets. Our aim was to construct a pipeline for synthetic dataset generation from aerial imagery and 3D models in Blender software. The dataset generation pipeline consists of seven steps and results in a wished number of images with bounding boxes in YOLO and coco formats. This synthetic dataset has been produced following the steps described in this pipeline. It consists of 5000 2048 × 2048 images with cars inserted into the roads and highways at the images without cars from all over the world. We believe that this dataset and the respective pipeline might be of great importance for vehicle detection, facilitating the customizability of the models to specific needs and context.

SUBMITTER: Oric M

PROVIDER: S-EPMC10875228 | biostudies-literature | 2024 Apr

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Synthetic car dataset for vehicle detection: Integrating aerial and satellite imagery.

Orić Mihaela M Galić Vlatko V Novoselnik Filip F

Data in brief 20240206

Vehicle detection is a very important aspect of computer vision application to aerial and satellite imagery, facilitating activities such as instance counting, velocity estimation, traffic predictions, etc. The feasibility of accurate vehicle detection often depends on limited training datasets, requiring a lot of manual work in collection and annotation tasks. Furthermore, there are no known publicly available datasets. Our aim was to construct a pipeline for synthetic dataset generation from a ...[more]

PMID: 38375146

Similar Datasets

Project description:Soil and water salinization has global impact on the sustainability of agricultural production, affecting the health and condition of staple crops and reducing potential yields. Identifying or developing salt-tolerant varieties of commercial crops is a potential pathway to enhance food and water security and deliver on the global demand for an increase in food supplies. Our study focuses on a phenotyping experiment that was designed to establish the influence of salinity stress on a diversity panel of the wild tomato species, Solanum pimpinellifolium. Here, we explore how unoccupied aerial vehicles (UAVs) equipped with both an optical and thermal infrared camera can be used to map and monitor plant temperature (Tp) changes in response to applied salinity stress. An object-based image analysis approach was developed to delineate individual tomato plants, while a green-red vegetation index derived from calibrated red, green, and blue (RGB) optical data allowed the discrimination of vegetation from the soil background. Tp was retrieved simultaneously from the co-mounted thermal camera, with Tp deviation from the ambient temperature and its change across time used as a potential indication of stress. Results showed that Tp differences between salt-treated and control plants were detectable across the five separate UAV campaigns undertaken during the field experiment. Using a simple statistical approach, we show that crop water stress index values greater than 0.36 indicated conditions of plant stress. The optimum period to collect UAV-based Tp for identifying plant stress was found between fruit formation and ripening. Preliminary results also indicate that UAV-based Tp may be used to detect plant stress before it is visually apparent, although further research with more frequent image collections and field observations is required. Our findings provide a tool to accelerate field phenotyping to identify salt-resistant germplasm and may allow farmers to alleviate yield losses through early detection of plant stress via management interventions.

Project description:Agricultural subsurface drainage systems are commonly installed on farmland to remove the excess water from poorly drained soils. Conventional methods for drainage mapping such as tile probes and trenching equipment are laborious, cause pipe damage, and are often inefficient to apply at large spatial scales. Knowledge of locations of an existing drainage network is crucial to understand the increased leaching and offsite release of drainage discharge and to retrofit the new drain lines within the existing drainage system. Recent technological developments in non-destructive techniques might provide a potential alternative solution. The objective of this study was to determine the suitability of unmanned aerial vehicle (UAV) imagery collected using three different cameras (visible-color, multispectral, and thermal infrared) and ground penetrating radar (GPR) for subsurface drainage mapping. Both the techniques are complementary in terms of their usage, applicability, and the properties they measure and were applied at four different sites in the Midwest USA. At Site-1, both the UAV imagery and GPR were equally successful across the entire field, while at Site-2, the UAV imagery was successful in one section of the field, and GPR proved to be useful in the other section where the UAV imagery failed to capture the drainage pipes' location. At Site-3, less to no success was observed in finding the drain lines using UAV imagery captured on bare ground conditions, whereas good success was achieved using GPR. Conversely, at Site-4, the UAV imagery was successful and GPR failed to capture the drainage pipes' location. Although UAV imagery seems to be an attractive solution for mapping agricultural subsurface drainage systems as it is cost-effective and can cover large field areas, the results suggest the usefulness of GPR to complement the former as both a mapping and validation technique. Hence, this case study compares and contrasts the suitability of both the methods, provides guidance on the optimal survey timing, and recommends their combined usage given both the technologies are available to deploy for drainage mapping purposes.

Project description:Monitoring the shoreline over time is essential to quickly identify and mitigate environmental issues such as coastal erosion. Monitoring using satellite images has two great advantages, i.e., global coverage and frequent measurement updates; but adequate methods are needed to extract shoreline information from such images. To this purpose, there are valuable non-supervised methods, but more recent research has concentrated on deep learning because of its greater potential in terms of generality, flexibility, and measurement accuracy, which, in contrast, derive from the information contained in large datasets of labeled samples. The first problem to solve, therefore, lies in obtaining large datasets suitable for this specific measurement problem, and this is a difficult task, typically requiring human analysis of a large number of images. In this article, we propose a technique to automatically create a dataset of labeled satellite images suitable for training machine learning models for shoreline detection. The method is based on the integration of data from satellite photos and data from certified, publicly accessible shoreline data. It involves several automatic processing steps, aimed at building the best possible dataset, with images including both sea and land regions, and correct labeling also in the presence of complicated water edges (which can be open or closed curves). The use of independently certified measurements for labeling the satellite images avoids the great work required to manually annotate them by visual inspection, as is done in other works in the literature. This is especially true when convoluted shorelines are considered. In addition, possible errors due to the subjective interpretation of satellite images are also eliminated. The method is developed and used specifically to build a new dataset of Sentinel-2 images, denoted SNOWED; but is applicable to different satellite images with trivial modifications. The accuracy of labels in SNOWED is directly determined by the uncertainty of the shoreline data used, which leads to sub-pixel errors in most cases. Furthermore, the quality of the SNOWED dataset is assessed through the visual comparison of a random sample of images and their corresponding labels, and its functionality is shown by training a neural model for sea-land segmentation.

Dataset Information

Synthetic car dataset for vehicle detection: Integrating aerial and satellite imagery.

Publications

Synthetic car dataset for vehicle detection: Integrating aerial and satellite imagery.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets