LEVIR-CD is a new large-scale remote sensing building Change Detection dataset. The introduced dataset would be a new benchmark for evaluating change detection (CD) algorithms, especially those based on deep learning. LEVIR-CD consists of 637 very high-resolution (VHR, 0.5m/pixel) Google Earth (GE) image patch pairs with a size of 1024 × 1024 pixels. These bitemporal images with time span of 5 to 14 years have significant land-use changes, especially the construction growth. LEVIR-CD covers various types of buildings, such as villa residences, tall apartments, small garages and large warehouses. Here, we focus on building-related changes, including the building growth (the change from soil/grass/hardened ground or building under construction to new build-up regions) and the building decline. These bitemporal images are annotated by remote sensing image interpretation experts using binary labels (1 for change and 0 for unchanged). Each sample in our dataset is annotated by one annotator and then double-checked by another to produce high-quality annotations. The fully annotated LEVIR-CD contains a total of 31,333 individual change-building instances.

The xBD dataset contains over 45,000KM2 of polygon labeled pre and post disaster imagery. The dataset provides the post-disaster imagery with transposed polygons from pre over the buildings, with damage classification labels. Source: xBD Image Source: Gupta et al

Million-AID is a large-scale benchmark dataset containing a million instances for RS scene classification. There are 51 semantic scene categories in Million-AID. And the scene categories are customized to match the land-use classification standards, which greatly enhance the practicability of the constructed Million-AID. Different form the existing scene classification datasets of which categories are organized with parallel or uncertain relationships, scene categories in Million-AID are organized with systematic relationship architecture, giving it superiority in management and scalability. Specifically, the scene categories in Million-AID are organized by the hierarchical category network of a three-level tree: 51 leaf nodes fall into 28 parent nodes at the second level which are grouped into 8 nodes at the first level, representing the 8 underlying scene categories of agriculture land, commercial land, industrial land, public service land, residential land, transportation land, unutilized land, and water area. The scene category network provides the dataset with excellent organization of relationship among different scene categories and also the property of scalability. The number of images in each scene category ranges from 2,000 to 45,000, endowing the dataset with the property of long tail distribution. Besides, Million-AID has superiorities over the existing scene classification datasets owing to its high spatial resolution, large scale, and global distribution. Source: DiRS: On Creating Benchmark Datasets for Remote Sensing Image Interpretation Image Source: https://captain-whu.github.io/DiRS/

AID is a new large-scale aerial image dataset, by collecting sample images from Google Earth imagery. Note that although the Google Earth images are post-processed using RGB renderings from the original optical aerial images, it has proven that there is no significant difference between the Google Earth images with the real optical aerial images even in the pixel-level land use/cover mapping. Thus, the Google Earth images can also be used as aerial images for evaluating scene classification algorithms. The new dataset is made up of the following 30 aerial scene types: airport, bare land, baseball field, beach, bridge, center, church, commercial, dense residential, desert, farmland, forest, industrial, meadow, medium residential, mountain, park, parking, playground, pond, port, railway station, resort, river, school, sparse residential, square, stadium, storage tanks and viaduct. All the images are labelled by the specialists in the field of remote sensing image interpretation, and some samples of each class are shown in Fig.1. In all, the AID dataset has a number of 10000 images within 30 classes. The images in AID are actually multi-source, as Google Earth images are from different remote imaging sensors. This brings more challenges for scene classification than the single source images like UC-Merced dataset. Moreover, all the sample images per each class in AID are carefully chosen from different countries and regions around the world, mainly in China, the United States, England, France, Italy, Japan, Germany, etc., and they are extracted at different time and seasons under different imaging conditions, which increases the intra-class diversities of the data.

Source: CHANGE DETECTION IN REMOTE SENSING IMAGES USING CONDITIONAL ADVERSARIAL NETWORKS

Gaofen Image Dataset (GID) is a large-scale land-cover dataset constructed with Gaofen-2 (GF-2) satellite images. This dataset has superiorities over the existing land-cover dataset because of its large coverage, wide distribution, and high spatial resolution. It contains 150 GF-2 images annotated at the pixel level for 5 categories: built-up, farmland, forest, meadow, and water.

SECOND is a well-annotated semantic change detection dataset. To ensure data diversity, we firstly collect 4662 pairs of aerial images from several platforms and sensors. These pairs of images are distributed over the cities such as Hangzhou, Chengdu, and Shanghai. Each image has size 512 x 512 and is annotated at the pixel level. The annotation of SECOND is carried out by an expert group of earth vision applications, which guarantees high label accuracy. For the change category in the SECOND dataset, we focus on 6 main land-cover classes, i.e. , non-vegetated ground surface, tree, low vegetation, water, buildings and playgrounds , that are frequently involved in natural and man-made geographical changes. It is worth noticing that, in the new dataset, non-vegetated ground surface ( n.v.g. surface for short) mainly corresponds to impervious surface and bare land. In summary, these 6 selected land-cover categories result in 30 common change categories (including non-change ). Through the random selection of image pairs, the SECOND reflects real distributions of land-cover categories when changes occur.

We manually edited an aerial and a satellite imagery dataset of building samples and named it a WHU building dataset. The aerial dataset consists of more than 220, 000 independent buildings extracted from aerial images with 0.075 m spatial resolution and 450 km2 covering in Christchurch, New Zealand. The satellite imagery dataset consists of two subsets. One of them is collected from cities over the world and from various remote sensing resources including QuickBird, Worldview series, IKONOS, ZY-3, etc. The other satellite building sub-dataset consists of 6 neighboring satellite images covering 550 km2 on East Asia with 2.7 m ground resolution.

SAMRS is a remote sensing segmentation dataset which provides object category, location, and instance information that can be used for semantic segmentation, instance segmentation, and object detection, either individually or in combination. Source: Scaling-up Remote Sensing Segmentation Dataset with Segment Anything Model Image Source: Scaling-up Remote Sensing Segmentation Dataset with Segment Anything Model

BANDON is a dataset for building change detection with off-nadir aerial images dataset, which is composed of off-Nadir image pairs of urban and rural areas. Overall, the BANDON dataset contains 2283 pairs of images, 2283 change labels,1891 BT-flows labels, 1891 pairs of segmentation labels, and 1891 pair of ST-offsets labels (test sets do not provide auxiliary annotations). Source:Detecting Building Changes with Off-Nadir Aerial Images Image Source: https://arxiv.org/pdf/2301.10922v1.pdf

The official dataset contains a training set (137 images), a validation set (4 images), and a testing set (10 images)

Multi-Sensor All Weather Mapping (MSAW) is a dataset and challenge, which features two collection modalities (both SAR and optical). The dataset and challenge focus on mapping and building footprint extraction using a combination of these data sources. MSAW covers 120 km^2 over multiple overlapping collects and is annotated with over 48,000 unique building footprints labels, enabling the creation and evaluation of mapping algorithms for multi-modal data. Source: SpaceNet 6: Multi-Sensor All Weather Mapping Dataset

For change detection tasks, current open-source datasets mainly focus on building extraction (e.g., WHU building dataset and LEVIR-CD dataset) (Chen and Shi, 2020; Ji et al., 2018) and urban development monitoring (e.g., SECOND dataset, Google dataset and CDD dataset) (Yang et al., 2022; Peng et al., 2021; Lebedev et al., 2018), whereas datasets for natural disaster monitoring have been seldom investigated. Therefore, we sought to present the GVLM dataset, the first large-scale and open-source VHR landslide mapping dataset. It includes $17$ bitemporal very-high-resolution imagery pairs with a spatial resolution of $0.59$ m acquired via Google Earth service. Each sub-dataset contains a pair of bitemporal images and the corresponding ground-truth map. The total coverage of the dataset is $163.77 km2$. The landslide sites in different geographical locations have various sizes, shapes, occurrence times, spatial distributions, phenology states, and land cover types, resulting in considerable spectral heterogeneity and intensity variations in the remote sensing imagery. The GVLM dataset can be used to develop and evaluate machine/deep learning models for change detection, semantic segmentation and landslide extraction.

WHU-RS19 is a set of satellite images exported from Google Earth, which provides high-resolution satellite images up to 0.5 m. Some samples of the database are displayed in the following picture. It contains 19 classes of meaningful scenes in high-resolution satellite imagery, including airport, beach, bridge, commercial, desert, farmland, footballfield, forest, industrial, meadow, mountain, park, parking, pond, port, railwaystation, residential, river, and viaduct. For each class, there are about 50 samples. It’s worth noticing that the image samples of the same class are collected from different regions in satellite images of different resolutions and then might have different scales, orientations and illuminations.

In this work, we propose a novel remote sensing dataset, FireRisk, consisting of 7 fire risk classes with a total of 91 872 labelled images for fire risk assessment. This remote sensing dataset is labelled with the fire risk classes supplied by the Wildfire Hazard Potential (WHP) raster dataset, and remote sensing images are collected using the National Agriculture Imagery Program (NAIP), a high-resolution remote sensing imagery program. On FireRisk, we present benchmark performance for supervised and self-supervised representations, with Masked Autoencoders (MAE) pre-trained on ImageNet1k achieving the highest classification accuracy, 65.29%.

The Five-Billion-Pixels dataset contains more than 5 billion labeled pixels of 150 high-resolution Gaofen-2 (4 m) satellite images, annotated in a 24-category system covering artificial-constructed, agricultural, and natural classes. It possesses the advantage of rich categories, large coverage, wide distribution, and high-spatial resolution, which well reflects the distributions of real-world ground objects and can benefit to different land cover related studies.

This dataset contains images from Sentinel-2 satellites taken before and after a wildfire. The ground truth masks are provided by the California Department of Forestry and Fire Protection and they are mapped on the images. The dataset is designed to do binary semantic segmentation of burned vs unburned areas.

A new benchmark dataset of webcam images, Photi-LakeIce, from multiple cameras and two different winters, along with pixel-wise ground truth annotations. Source: Lake Ice Monitoring with Webcams and Crowd-Sourced Images

The DRIFT dataset includes 25k image patches collected in five European countries sourced from aerial and nanosatellite image archives. Each image patch is associated with three target variables to predict: 1. Canopy height: average height value for pixels containing woody vegetation. 2. Tree count: number of overstory (visible from an overhead perspective) trees in the images. 3. Tree cover fraction: percentage of the image being covered by overstory tree crowns. The DRIFT dataset includes significant shifts between label and visual distributions due to sensor and area differences. Furthermore, vegetation tends to grow to fit the local climate, therefore introducing concept drift in the data: same tree species may appear differently in different subsets. The label distribution also varies among different subsets (countries). The dataset is a good choice for: * image-level regression * domain adaption for regression * remote sensing for forest applications

The GDIT Aerial Airport dataset consists of aerial images containing instances of parked airplanes. All plane types have been grouped into a single classification named "airplane".

The HRPlanesv2 dataset contains 2120 VHR Google Earth images. To further improve experiment results, images of airports from many different regions with various uses (civil/military/joint) selected and labeled. A total of 14,335 aircrafts have been labelled. Each image is stored as a ".jpg" file of size 4800 x 2703 pixels and each label is stored as YOLO ".txt" format. Dataset has been split in three parts as 70% train, %20 validation and test. The aircrafts in the images in the train and validation datasets have a percentage of 80 or more in size. Link: https://github.com/dilsadunsal/HRPlanesv2-Data-Set

A high-resolution multi-sensor remote sensing scene classification dataset, appropriate for training and evaluating image classification models in the remote sensing domain. The dataset consists of 8400 overhead scenes, each covered by Airbus Pléiades, Airbus SPOT, and USDA NAIP imagery. The scenes are classified into 12 OpenStreetMap categories: * man_made=bridge * man_made=breakwater * building=farm * power=substation * leisure=stadium * leisure=golf_course * waterway=dam * landuse=quarry * landuse=farmland * landuse=forest * natural=water * natural=bare_rock

This dataset includes 3D point-cloud and 2D imagery from a flash LiDAR...

UV6K is a high-resolution remote sensing urban vehicle segmentation dataset. - Images: 6,313 - Vehicle: 245,141 - Resolution: 0.1m - Image Size: 1024x1024

The dataset comprises patches of size 512x512 pixels collected from Sentinel-2 L2A satellite mission. All reported forest fires are located in California. For each area of interest, two images are provided: pre-fire acquisition and post-fire acquisition. Each image is composed of 12 different channels, collecting information from the visible spectrum, infrared and ultrablue. The dataset is split into: - train set - validation set - hidden test set, for which ground truth labels are not disclosed. The hdf5 file is structured in this way: root | |- uuid_0: {"post_fire", "pre_fire", "mask"} | | |- uuid_1: {"post_fire", "pre_fire", "mask"} | ... Each uuid have associated an attribute called fold. Dataset can be downloaded from here. You can find a script that can be used to load data. Additional optional data are available here. If you need more data, contact us.

MapAI: Precision in Building Segmentation Dataset The dataset comprises 7500 training images and 1500 validation images from Denmark. The test dataset is split into two tasks, where the first task (1368 images) is to segment the buildings only using aerial images. In contrast, the second task (978 images) allows using aerial images and lidar data. All data samples have a resolution of 500x500. The aerial images are RGB images, while the lidar data are rasterized. The ground truth masks have two classes, building, and background. Source: MapAI: Precision in Building Segmentaiton Image Source: Kartai

RTI International (RTI) generated 2,611 labeled point locations representing 19 different land cover types, clustered in 5 distinct agroecological zones within Rwanda. These land cover types were reduced to three crop types (Banana, Maize, and Legume), two additional non-crop land cover types (Forest and Structure), and a catch-all Other land cover type to provide training/evaluation data for a crop classification model. Each point is attributed with its latitude and longitude, the land cover type, and the degree of confidence the labeler had when classifying the point location. For each location there are also three corresponding image chips (4.5 m x 4.5 m in size) with the point id as part of the image name. Each image contains a P1, P2, or P3 designation in the name, indicating the time period. P1 corresponds to December 2018, P2 corresponds to January 2019, and P3 corresponds to February 2019. These data were used in the development of research documented in greater detail in “Deep Neural Networks and Transfer Learning for Food Crop Identification in UAV Images” (Chew et al., 2020).