-This page lists resources on deep learning applied to satellite and aerial imagery. For regular updates 👇
-
+This page lists resources on deep learning applied to satellite and aerial imagery. For regular updates & videos 👇
# 🖥️ [satellite-image-deep-learning.com](https://www.satellite-image-deep-learning.com/)
*Are you looking for a remote sensing dataset, or want to know about training and deploying models?*
-
# 📺 [Video: the remote-sensing-datasets repository](https://www.satellite-image-deep-learning.com/p/datasets-for-machine-and-deep-learning)
-*Do you want expert advice on your satellite & aerial imagery project?*
-
+*Do you want expert advice on your satellite or aerial imagery project?*
# 💬 [Arrange a 1:1 with me](https://robmarkcole.com/consulting.html)
-*Do you want to support my open source work?*
-
-# 🎖️ [Sponsor me on Github](https://github.com/sponsors/robmarkcole)
-
-**How to use this repository:** if you know exactly what you are looking for (e.g. a dataset name) you can `Control+F` to search for it in the page. Otherwise refer to the table of contents below, and search within the relevant section. Material that is suitable for getting started with a topic is tagged with `BEGINNER`, which can also be searched.
-
-# Table of contents
-* [Techniques](https://github.com/robmarkcole/satellite-image-deep-learning#techniques)
- * [Classification](https://github.com/robmarkcole/satellite-image-deep-learning#Classification)
- * [Segmentation](https://github.com/robmarkcole/satellite-image-deep-learning#Segmentation)
- * [Land use & land cover](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---land-use--land-cover)
- * [Vegetation, crops & crop boundaries](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---vegetation-crops--crop-boundaries)
- * [Water, coastlines & floods](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---water-coastlines--floods)
- * [Fire, smoke & burn areas](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---fire-smoke--burn-areas)
- * [Landslides](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---landslides)
- * [Glaciers](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---glaciers)
- * [Other environmental](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---other-environmental)
- * [Roads](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---roads)
- * [Buildings & rooftops](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---buildings--rooftops)
- * [Solar panels](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---solar-panels)
- * [Other manmade](https://github.com/robmarkcole/satellite-image-deep-learning#segmentation---other-manmade)
- * [Instance segmentation](https://github.com/robmarkcole/satellite-image-deep-learning#instance-segmentation)
- * [Panoptic segmentation](https://github.com/robmarkcole/satellite-image-deep-learning#panoptic-segmentation)
- * [Object detection](https://github.com/robmarkcole/satellite-image-deep-learning#object-detection)
- * [Object detection with rotated bounding boxes](https://github.com/robmarkcole/satellite-image-deep-learning#object-detection-with-rotated-bounding-boxes)
- * [Object detection enhanced by super resolution](https://github.com/robmarkcole/satellite-image-deep-learning#object-detection-enhanced-by-super-resolution)
- * [Salient object detection](https://github.com/robmarkcole/satellite-image-deep-learning#salient-object-detection)
- * [Buildings, rooftops & solar panels](https://github.com/robmarkcole/satellite-image-deep-learning#object-detection---buildings-rooftops--solar-panels)
- * [Ships & boats](https://github.com/robmarkcole/satellite-image-deep-learning#object-detection---ships--boats)
- * [Cars, vehicles & trains](https://github.com/robmarkcole/satellite-image-deep-learning#object-detection---cars-vehicles--trains)
- * [Planes & aircraft](https://github.com/robmarkcole/satellite-image-deep-learning#object-detection---planes--aircraft)
- * [Infrastructure & utilities](https://github.com/robmarkcole/satellite-image-deep-learning#object-detection---infrastructure--utilities)
- * [Oil storage tank detection](https://github.com/robmarkcole/satellite-image-deep-learning#object-detection---oil-storage-tank-detection)
- * [Animals](https://github.com/robmarkcole/satellite-image-deep-learning#object-detection---animals)
- * [Object counting](https://github.com/robmarkcole/satellite-image-deep-learning#object-counting)
- * [Regression](https://github.com/robmarkcole/satellite-image-deep-learning#regression)
- * [Cloud detection & removal](https://github.com/robmarkcole/satellite-image-deep-learning#cloud-detection--removal)
- * [Change detection](https://github.com/robmarkcole/satellite-image-deep-learning#change-detection)
- * [Time series](https://github.com/robmarkcole/satellite-image-deep-learning#time-series)
- * [Crop yield](https://github.com/robmarkcole/satellite-image-deep-learning#crop-yield)
- * [Wealth and economic activity](https://github.com/robmarkcole/satellite-image-deep-learning#wealth-and-economic-activity)
- * [Disaster response](https://github.com/robmarkcole/satellite-image-deep-learning#disaster-response)
- * [Super-resolution](https://github.com/robmarkcole/satellite-image-deep-learning#super-resolution)
- * [Pansharpening](https://github.com/robmarkcole/satellite-image-deep-learning#pansharpening)
- * [Image-to-image translation](https://github.com/robmarkcole/satellite-image-deep-learning#image-to-image-translation)
- * [GANS](https://github.com/robmarkcole/satellite-image-deep-learning#gans)
- * [Transformers](https://github.com/robmarkcole/satellite-image-deep-learning#transformers)
- * [Adversarial ML](https://github.com/robmarkcole/satellite-image-deep-learning#adversarial-ml)
- * [Autoencoders, dimensionality reduction, image embeddings & similarity search](https://github.com/robmarkcole/satellite-image-deep-learning#autoencoders-dimensionality-reduction-image-embeddings--similarity-search)
- * [Image retreival](https://github.com/robmarkcole/satellite-image-deep-learning#image-retreival)
- * [Image Captioning & Visual Question Answering](https://github.com/robmarkcole/satellite-image-deep-learning#image-captioning--visual-question-answering)
- * [Mixed data learning](https://github.com/robmarkcole/satellite-image-deep-learning#mixed-data-learning)
- * [Few-shot learning](https://github.com/robmarkcole/satellite-image-deep-learning#few-shot-learning)
- * [Self-supervised, unsupervised & contrastive learning](https://github.com/robmarkcole/satellite-image-deep-learning#self-supervised-unsupervised--contrastive-learning)
- * [Weakly & semi-supervised learning](https://github.com/robmarkcole/satellite-image-deep-learning#weakly--semi-supervised-learning)
- * [Active learning](https://github.com/robmarkcole/satellite-image-deep-learning#active-learning)
- * [Federated learning](https://github.com/robmarkcole/satellite-image-deep-learning#federated-learning)
- * [Image registration](https://github.com/robmarkcole/satellite-image-deep-learning#image-registration)
- * [Data fusion](https://github.com/robmarkcole/satellite-image-deep-learning#data-fusion)
- * [Terrain mapping, Disparity Estimation, Lidar, DEMs & NeRF](https://github.com/robmarkcole/satellite-image-deep-learning#terrain-mapping-disparity-estimation-lidar-dems--nerf)
- * [Thermal Infrared](https://github.com/robmarkcole/satellite-image-deep-learning#thermal-infrared)
- * [SAR](https://github.com/robmarkcole/satellite-image-deep-learning#sar)
- * [NDVI - vegetation index](https://github.com/robmarkcole/satellite-image-deep-learning#ndvi---vegetation-index)
- * [General image quality](https://github.com/robmarkcole/satellite-image-deep-learning#general-image-quality)
- * [Synthetic data](https://github.com/robmarkcole/satellite-image-deep-learning#synthetic-data)
-* [Deep learning projects & frameworks](https://github.com/robmarkcole/satellite-image-deep-learning#deep-learning-projects--frameworks)
-* [Neural nets in space](https://github.com/robmarkcole/satellite-image-deep-learning#neural-nets-in-space)
-* [Courses](https://github.com/robmarkcole/satellite-image-deep-learning#courses)
-* [Podcasts](https://github.com/robmarkcole/satellite-image-deep-learning#podcasts)
-* [Newsletters](https://github.com/robmarkcole/satellite-image-deep-learning#newsletters)
-* [Books](https://github.com/robmarkcole/satellite-image-deep-learning#books)
+*Do you want to support this resource?*
+# 🎖️ [See my Github sponsors](https://github.com/sponsors/robmarkcole)
+
+**How to use this repository:** if you know exactly what you are looking for (e.g. you have the paper name) you can `Control+F` to search for it in this page (or search in the raw markdown). Note that material that is suitable for getting started with a topic is tagged with BEGINNER, which can also be searched.
# Techniques
-This section explores the different deep and machine learning (ML) techniques applied to common problems in satellite imagery analysis. Good background reading is [Deep learning in remote sensing applications: A meta-analysis and review](https://www.iges.or.jp/en/publication_documents/pub/peer/en/6898/Ma+et+al+2019.pdf)
+This section lists techniques applied to satellite & aerial imagery. Click the ▶️ arrow to expand the section
-## Classification
++ +### Classification Classification is the task of assigning a label to an image, e.g. this is an image of a forest. The more complex case is applying multiple labels to an image. This approach of image level classification is not to be confused with pixel-level classification which is called semantic segmentation. Read more in my post [A brief introduction to satellite image classification with neural networks](https://www.satellite-image-deep-learning.com/p/a-brief-introduction-to-satellite) * Land classification on Sentinel 2 data using a [simple sklearn cluster algorithm](https://github.com/acgeospatial/Satellite_Imagery_Python/blob/master/Clustering_KMeans-Sentinel2.ipynb) or [deep learning CNN](https://towardsdatascience.com/land-use-land-cover-classification-with-deep-learning-9a5041095ddb) `BEGINNER` * Land Use Classification on Merced dataset using CNN [in Keras](https://github.com/tavgreen/landuse_classification) @@ -163,7 +96,13 @@ or [fastai](https://medium.com/spatial-data-science/deep-learning-for-geospatial * [satellite-deforestation](https://github.com/drewhibbard/satellite-deforestation) -> Using Satellite Imagery to Identify the Leading Indicators of Deforestation, applied to the Kaggle Challenge Understanding the Amazon from Space * [RSMLC](https://github.com/marjanstoimchev/RSMLC) -> code for 2023 [paper](https://www.mdpi.com/2072-4292/15/2/538): Deep Network Architectures as Feature Extractors for Multi-Label Classification of Remote Sensing Images -## Segmentation +
++ +### Segmentation Segmentation will assign a class label to each **pixel** in an image. Single class models are often trained for road or building segmentation, with multi class for land use/crop type classification. **Note** that many articles which refer to 'hyperspectral land classification' are actually describing semantic segmentation. Note that cloud detection can be addressed with semantic segmentation and has its own section [Cloud detection & removal](https://github.com/robmarkcole/satellite-image-deep-learning#cloud-detection--removal). Read more about segmentation in my post [A brief introduction to satellite image segmentation with neural networks](https://www.satellite-image-deep-learning.com/p/a-brief-introduction-to-satellite-365) * [awesome-satellite-images-segmentation](https://github.com/mrgloom/awesome-semantic-segmentation#satellite-images-segmentation) * [Satellite Image Segmentation: a Workflow with U-Net](https://medium.com/vooban-ai/satellite-image-segmentation-a-workflow-with-u-net-7ff992b2a56e) is a decent intro article `BEGINNER` @@ -560,7 +499,13 @@ In instance segmentation, each individual 'instance' of a segmented area is give * [BSB-Aerial-Dataset](https://github.com/osmarluiz/BSB-Aerial-Dataset) -> an example on how to use Detectron2's Panoptic-FPN in the BSB Aerial Dataset * [utae-paps](https://github.com/VSainteuf/utae-paps) -> PyTorch implementation of U-TAE and PaPs for satellite image time series panoptic segmentation -## Object detection +
++ +### Object detection Onject detection is the task of placing a box around the bounds of an object (i.e. creating a bounding box). Typical use cases are detecting vehicles, aircraft & ships. A good introduction to the challenge of performing object detection on aerial imagery is given in [this paper](https://arxiv.org/abs/1902.06042v2). In summary, images are large and objects may comprise only a few pixels, easily confused with random features in background. For the same reason, object detection datasets are inherently imbalanced, since the area of background typically dominates over the area of the objects to be detected. In general object detection performs well on large objects, and gets increasingly difficult as the objects get smaller & more densely packed. Model accuracy falls off rapidly as image resolution degrades, so it is common for object detection to use very high resolution imagery, e.g. 30cm RGB. A particular characteristic of aerial images is that objects can be oriented in any direction, so using rotated bounding boxes which align with the object can be crucial for extracting measurements of the length and width of an object. * [Object detection on Satellite Imagery using RetinaNet](https://medium.com/@ije_good/object-detection-on-satellite-imagery-using-retinanet-part-1-training-e589975afbd5) -> using the Kaggle Swimming Pool and Car Detection dataset `BEGINNER` * [Tackling the Small Object Problem in Object Detection](https://blog.roboflow.com/tackling-the-small-object-problem-in-object-detection) `BEGINNER` @@ -826,14 +771,26 @@ When the object count, but not its shape is required, U-net can be used to treat * [PSGCNet](https://github.com/gaoguangshuai/PSGCNet) -> code for 2022 [paper](https://arxiv.org/abs/2012.03597): PSGCNet: A Pyramidal Scale and Global Context Guided Network for Dense Object Counting in Remote Sensing Images * [psgcnet](https://github.com/gaoguangshuai/psgcnet) -> code for 2022 [paper](https://ieeexplore.ieee.org/abstract/document/9720990): PSGCNet: A Pyramidal Scale and Global Context Guided Network for Dense Object Counting in Remote-Sensing Images -## Regression +
++ +### Regression * [python-windspeed](https://github.com/h-fuzzy-logic/python-windspeed) -> Predicting windspeed of hurricanes from satellite images, uses CNN regression in keras * [hurricane-wind-speed-cnn](https://github.com/23ccozad/hurricane-wind-speed-cnn) -> Predicting windspeed of hurricanes from satellite images, uses CNN regression in keras * [GEDI-BDL](https://github.com/langnico/GEDI-BDL) -> code for paper: Global canopy height regression and uncertainty estimation from GEDI LIDAR waveforms with deep ensembles * [Traffic density estimation as a regression problem instead of object detection](https://omdena.com/blog/ai-road-safety/) -> inspired by [this paper](https://ieeexplore.ieee.org/document/8916990) * [OpticalWaveGauging_DNN](https://github.com/OpticalWaveGauging/OpticalWaveGauging_DNN) -> code for 2020 [paper](https://www.sciencedirect.com/science/article/abs/pii/S0378383919301243): Optical wave gauging using deep neural networks -## Cloud detection & removal +
++ +### Cloud detection & removal There are many algorithms that use band math to detect clouds, but the deep learning approach is to use semantic segmentation * [CloudSEN12](https://github.com/cloudsen12) -> Sentinel 2 cloud dataset with a [varierty of models here](https://github.com/cloudsen12/models) * See section [Kaggle - Understanding Clouds from Satellite Images](https://github.com/robmarkcole/satellite-image-deep-learning#kaggle---understanding-clouds-from-satellite-images) @@ -896,7 +853,13 @@ There are many algorithms that use band math to detect clouds, but the deep lear * [cloud-detection-venus](https://github.com/pesekon2/cloud-detection-venus) -> code for 2022 [paper](https://www.mdpi.com/2072-4292/14/20/5210): Using Convolutional Neural Networks for Cloud Detection on VENμS Images over Multiple Land-Cover Types * [explaining_cloud_effects](https://github.com/JakobCode/explaining_cloud_effects) -> code for 2022 [paper](https://ieeexplore.ieee.org/abstract/document/9956865): Explaining the Effects of Clouds on Remote Sensing Scene Classification -## Change detection +
++ +### Change detection Generally speaking, change detection methods are applied to a pair of images to generate a mask of change, e.g. of buildings damaged in a disaster. Note, clouds & shadows change often too..! * [awesome-remote-sensing-change-detection](https://github.com/wenhwu/awesome-remote-sensing-change-detection) lists many datasets and publications * [Change-Detection-Review](https://github.com/MinZHANG-WHU/Change-Detection-Review) -> A review of change detection methods, including code and open data sets for deep learning @@ -993,7 +956,13 @@ Generally speaking, change detection methods are applied to a pair of images to * [A-loss-function-for-change-detection](https://github.com/Chuan-shanjia/A-loss-function-for-change-detection) -> code for 2022 [paper](https://ieeexplore.ieee.org/document/9884382): UAL: Unchanged Area Loss-Function for Change Detection Networks * [IEEE_TGRS_SSTFormer](https://github.com/yanhengwang-heu/IEEE_TGRS_SSTFormer) -> code for 2022 [paper](https://ieeexplore.ieee.org/abstract/document/9870837): Spectral–Spatial–Temporal Transformers for Hyperspectral Image Change Detection -## Time series +
++ +### Time series More general than change detection, time series observations can be used for applications including improving the accuracy of crop classification, or predicting future patterns & events. Crop yield is very typically application and has its own section below * [CropDetectionDL](https://github.com/karimmamer/CropDetectionDL) -> using GRU-net, First place solution for Crop Detection from Satellite Imagery competition organized by CV4A workshop at ICLR 2020 * [LANDSAT Time Series Analysis for Multi-temporal Land Cover Classification using Random Forest](https://github.com/agr-ayush/Landsat-Time-Series-Analysis-for-Multi-Temporal-Land-Cover-Classification) @@ -1020,7 +989,13 @@ More general than change detection, time series observations can be used for app * [satellite_image_forecasting](https://github.com/rudolfwilliam/satellite_image_forecasting) -> predict future satellite images from past ones using features such as precipitation and elevation maps. Entry for the [EarthNet2021](https://www.earthnet.tech/) challenge * [Deep Learning for Cloud Gap-Filling on Normalized Difference Vegetation Index using Sentinel Time-Series](https://github.com/Agri-Hub/Deep-Learning-for-Cloud-Gap-Filling-on-Normalized-Difference-Vegetation-Index) -> A CNN-RNN based model that identifies correlations between optical and SAR data and exports dense Normalized Difference Vegetation Index (NDVI) time-series of a static 6-day time resolution and can be used for Events Detection tasks -## Crop yield +
++ +### Crop yield * [Crop yield Prediction with Deep Learning](https://github.com/JiaxuanYou/crop_yield_prediction) -> code for the paper Deep Gaussian Process for Crop Yield Prediction Based on Remote Sensing Data * [Deep-Transfer-Learning-Crop-Yield-Prediction](https://github.com/sustainlab-group/Deep-Transfer-Learning-Crop-Yield-Prediction) * [Crop-Yield-Prediction-using-ML](https://github.com/VaibhavSaini19/Crop-Yield-Prediction-using-ML) -> A simple Web application developed in order to provide the farmers/users an approximation on how much amount of crop yield will be produced depending upon the given input @@ -1036,7 +1011,13 @@ More general than change detection, time series observations can be used for app * [SPACY](https://github.com/rlee360/PLaTYPI) -> Satellite Prediction of Aggregate Corn Yield * [cropyieldArticle](https://github.com/myliheik/cropyieldArticle) -> code for 2022 [paper](https://www.mdpi.com/2072-4292/14/17/4193): Scalable Crop Yield Prediction with Sentinel-2 Time Series and Temporal Convolutional Network -## Wealth and economic activity +
++ +### Wealth and economic activity The goal is to predict economic activity from satellite imagery rather than conducting labour intensive ground surveys * [Using publicly available satellite imagery and deep learning to understand economic well-being in Africa, Nature Comms 22 May 2020](https://www.nature.com/articles/s41467-020-16185-w) -> Used CNN on Ladsat imagery (night & day) to predict asset wealth of African villages * [Combining Satellite Imagery and machine learning to predict poverty](https://towardsdatascience.com/combining-satellite-imagery-and-machine-learning-to-predict-poverty-884e0e200969) -> review article @@ -1062,7 +1043,14 @@ The goal is to predict economic activity from satellite imagery rather than cond * [uganda-poverty-project](https://github.com/vinceranga/uganda-poverty-project) -> use through Object Detection on high-resolution satellite imagery to identify indicators of poverty and economic inequality within Uganda * [Predicting_Energy_Consumption_With_Convolutional_Neural_Networks](https://github.com/healdz/Predicting_Energy_Consumption_With_Convolutional_Neural_Networks) -## Disaster response +
++ + +### Disaster response Also checkout the sections on change detection and water/fire/building segmentation * [DisaVu](https://github.com/SrzStephen/DisaVu) -> combines building & damage detection and provides an app for viewing predictions * [Soteria](https://github.com/Soteria-ai/Soteria) -> uses machine learning with satellite imagery to map natural disaster impacts for faster emergency response @@ -1085,7 +1073,13 @@ Also checkout the sections on change detection and water/fire/building segmentat * [shackleton](https://github.com/avanetten/shackleton) -> leverages remote sensing imagery and machine learning techniques to provide insights into various transportation and evacuation scenarios in an interactive dashboard that conducts real-time computation * [ai-vegetation-fuel](https://github.com/ecmwf-projects/ai-vegetation-fuel) -> Predicting Fuel Load from earth observation data using Machine Learning, using LightGBM & CatBoost -## Super-resolution +
++ +### Super-resolution Super-resolution attempts to enhance the resolution of an imaging system, and can be applied as a pre-processing step to improve the detection of small objects or boundaries. Its use is controversial since it can introduce artefacts at the same rate as real features. These techniques are generally grouped into single image super resolution (SISR) **or** a multi image super resolution (MISR) * [The value of super resolution — real world use case](https://medium.com/sentinel-hub/the-value-of-super-resolution-real-world-use-case-2ba811f4cd7f) -> Medium article on parcel boundary detection with super-resolved satellite imagery * [Super-Resolution on Satellite Imagery using Deep Learning](https://medium.com/the-downlinq/super-resolution-on-satellite-imagery-using-deep-learning-part-1-ec5c5cd3cd2) -> Nov 2016 blog post by CosmiQ Works with a nice introduction to the topic. Proposes and demonstrates a new architecture with perturbation layers with practical guidance on the methodology and [code](https://github.com/CosmiQ/super-resolution). [Three part series](https://medium.com/the-downlinq/super-resolution-on-satellite-imagery-using-deep-learning-part-3-2e2f61eee1d3) @@ -1157,7 +1151,13 @@ Note that nearly all the MISR publications resulted from the [PROBA-V Super Reso * [worldstrat](https://github.com/worldstrat/worldstrat) -> SISR and MISR implementations of SRCNN * [MISR-GRU](https://github.com/rarefin/MISR-GRU) -> Pytorch implementation of MISR-GRU, a deep neural network for multi image super-resolution (MISR), for ProbaV Super Resolution Competition -## Pansharpening +
++ +### Pansharpening Image fusion of low res multispectral with high res pan band. * Several algorithms described [in the ArcGIS docs](http://desktop.arcgis.com/en/arcmap/10.3/manage-data/raster-and-images/fundamentals-of-panchromatic-sharpening.htm), with the simplest being taking the mean of the pan and RGB pixel value. * For into to classical methods [see this notebook](http://nbviewer.jupyter.org/github/HyperionAnalytics/PyDataNYC2014/blob/master/panchromatic_sharpening.ipynb) and [this kaggle kernel](https://www.kaggle.com/resolut/panchromatic-sharpening) @@ -1192,7 +1192,13 @@ Image fusion of low res multispectral with high res pan band. * [PanFormer](https://github.com/zhysora/PanFormer) -> code for 2022 [paper](https://arxiv.org/abs/2203.02916): PanFormer: a Transformer Based Model for Pan-sharpening * [Pansharpening](https://github.com/nithin-gr/Pansharpening) -> code for 2021 [paper](https://www.researchgate.net/publication/356974466_Pansformers_Transformer-Based_Self-Attention_Network_for_Pansharpening): Pansformers: Transformer-Based Self-Attention Network for Pansharpening -## Image-to-image translation +
++ +### Image-to-image translation Translate images e.g. from SAR to RGB. * [How to Develop a Pix2Pix GAN for Image-to-Image Translation](https://machinelearningmastery.com/how-to-develop-a-pix2pix-gan-for-image-to-image-translation/) -> how to develop a Pix2Pix model for translating satellite photographs to Google map images. A good intro to GANS * [A growing problem of ‘deepfake geography’: How AI falsifies satellite images](https://www.washington.edu/news/2021/04/21/a-growing-problem-of-deepfake-geography-how-ai-falsifies-satellite-images/) @@ -1211,7 +1217,13 @@ Translate images e.g. from SAR to RGB. * [sat_to_map](https://github.com/shagunuppal/sat_to_map) -> Learning mappings to generate city maps images from corresponding satellite images * [pix2pix-GANs](https://github.com/shashi7679/pix2pix-GANs) -> Generate Map using Satellite Image & PyTorch -## GANS +
++ +### GANS GANS are famously used for generating synthetic data, see the section [Synthetic data](https://github.com/robmarkcole/satellite-image-deep-learning#synthetic-data) * [Anomaly Detection on Mars using a GAN](https://omdena.com/projects/anomaly-detection-mars/) * [Using Generative Adversarial Networks to Address Scarcity of Geospatial Training Data](https://medium.com/radiant-earth-insights/using-generative-adversarial-networks-to-address-scarcity-of-geospatial-training-data-e61cacec986e) -> GAN perform better than CNN in segmenting land cover classes outside of the training dataset (article, no code) @@ -1232,19 +1244,38 @@ GANS are famously used for generating synthetic data, see the section [Synthetic * [PSGan-Family](https://github.com/zhysora/PSGan-Family) -> code for 2020 [paper](https://ieeexplore.ieee.org/document/9306912): PSGAN: A Generative Adversarial Network for Remote Sensing Image Pan-Sharpening * [Satellite Image Augmetation with GANs](https://github.com/Oarowolo11/11785-Project) -> code for 2022 [paper](https://arxiv.org/abs/2207.14580): Image Augmentation for Satellite Images -## Transformers +
++ +### Transformers * [Transformer-in-Remote-Sensing](https://github.com/VIROBO-15/Transformer-in-Remote-Sensing) -> code for 2022 [paper](https://arxiv.org/abs/2209.01206): Transformers in Remote Sensing: A Survey * [Remote-Sensing-RVSA](https://github.com/ViTAE-Transformer/Remote-Sensing-RVSA) -> code for 2022 [paper](https://arxiv.org/abs/2208.03987): Advancing Plain Vision Transformer Towards Remote Sensing Foundation Model * [SatViT](https://github.com/antofuller/SatViT) -> self-supervised training of multispectral optical and SAR vision transformers * [UDA_for_RS](https://github.com/Levantespot/UDA_for_RS) -> code for 2022 [paper](https://www.mdpi.com/2072-4292/14/19/4942): Unsupervised Domain Adaptation for Remote Sensing Semantic Segmentation with Transformer -## Adversarial ML +
++ +### Adversarial ML Efforts to detect falsified images & deepfakes. Also checkout [Synthetic data](https://github.com/robmarkcole/satellite-image-deep-learning#synthetic-data) * [UAE-RS](https://github.com/YonghaoXu/UAE-RS) -> dataset that provides black-box adversarial samples in the remote sensing field * [PSGAN](https://github.com/xuxiangsun/PSGAN) -> code for paper: Perturbation Seeking Generative Adversarial Networks: A Defense Framework for Remote Sensing Image Scene Classification * [SACNet](https://github.com/YonghaoXu/SACNet) -> code for 2021 [paper](https://ieeexplore.ieee.org/document/9573256): Self-Attention Context Network: Addressing the Threat of Adversarial Attacks for Hyperspectral Image Classification -## Autoencoders, dimensionality reduction, image embeddings & similarity search +
++ + +### Autoencoders, dimensionality reduction, image embeddings & similarity search * [Autoencoders & their Application in Remote Sensing](https://towardsdatascience.com/autoencoders-their-application-in-remote-sensing-95f6e2bc88f) -> intro article and example use case applied to SAR data for land classification * [LEt-SNE](https://github.com/meghshukla/LEt-SNE) -> Dimensionality Reduction and visualization technique that compensates for the curse of dimensionality * [AutoEncoders for Land Cover Classification of Hyperspectral Images](https://towardsdatascience.com/autoencoders-for-land-cover-classification-of-hyperspectral-images-part-1-c3c847ebc69b) -> An autoencoder nerual net is used to reduce 103 band data to 60 features (dimensionality reduction), keras. Also read [part 2](https://syamkakarla.medium.com/auto-encoders-for-land-cover-classification-in-hyperspectral-images-part-2-f8978d443d6d) which implements K-NNC, SVM and Gradient Boosting @@ -1281,7 +1312,14 @@ Efforts to detect falsified images & deepfakes. Also checkout [Synthetic data](h * [Train SimSiam on Satellite Images](https://docs.lightly.ai/tutorials/package/tutorial_simsiam_esa.html) using lightly.ai to generate embeddings that can be used for data exploration and understanding * [Airbus_SDC_dup](https://github.com/WillieMaddox/Airbus_SDC_dup) -> Project focused on detecting duplicate regions of overlapping satellite imagery. Applied to Airbus ship detection dataset -## Image retreival +
++ + +### Image retreival * [Demo_AHCL_for_TGRS2022](https://github.com/weiweisong415/Demo_AHCL_for_TGRS2022) -> code for 2022 paper: Asymmetric Hash Code Learning (AHCL) for remote sensing image retreival * [GaLR](https://github.com/xiaoyuan1996/GaLR) -> code for 2022 [paper](https://ieeexplore.ieee.org/abstract/document/9745546): Remote Sensing Cross-Modal Text-Image Retrieval Based on Global and Local Information * [retrievalSystem](https://github.com/xiaoyuan1996/retrievalSystem) -> cross-modal image retrieval system @@ -1293,7 +1331,13 @@ Efforts to detect falsified images & deepfakes. Also checkout [Synthetic data](h * [HydroViet_VOR](https://github.com/lannguyen0910/HydroViet_VOR) -> Object Retrieval in satellite images with Triplet Network * [AMFMN](https://github.com/AICyberTeam/AMFMN) -> code for 2021 [paper](https://ieeexplore.ieee.org/document/9437331): Exploring a Fine-Grained Multiscale Method for Cross-Modal Remote Sensing Image Retrieval -## Image Captioning & Visual Question Answering +
++ +### Image Captioning & Visual Question Answering * See the section [Image captioning datasets](https://github.com/robmarkcole/satellite-image-deep-learning#image-captioning-datasets) * [remote-sensing-image-caption](https://github.com/TalentBoy2333/remote-sensing-image-caption) -> image classification and image caption by PyTorch * [Fine tuning CLIP with Remote Sensing (Satellite) images and captions](https://huggingface.co/blog/fine-tune-clip-rsicd) -> fine tuning CLIP on the [RSICD](https://github.com/201528014227051/RSICD_optimal) image captioning dataset, to enable querying large catalogues in natural language. With [repo](https://github.com/arampacha/CLIP-rsicd), uses 🤗 @@ -1308,7 +1352,13 @@ Efforts to detect falsified images & deepfakes. Also checkout [Synthetic data](h * [MetaCaptioning](https://github.com/QiaoqiaoYang/MetaCaptioning) -> code for 2022 [paper](https://www.sciencedirect.com/science/article/abs/pii/S0924271622000351): Meta captioning: A meta learning based remote sensing image captioning framework * [Transformer-for-image-captioning](https://github.com/RicRicci22/Transformer-for-image-captioning) -> a transformer for image captioning, trained on the UCM dataset -## Mixed data learning +
++ +### Mixed data learning These techniques combine multiple data types, e.g. imagery and text data. * [Predicting the locations of traffic accidents with satellite imagery and convolutional neural networks](https://towardsdatascience.com/teaching-a-neural-network-to-see-roads-74bff240c3e5) -> Combining satellite imagery and structured data to predict the location of traffic accidents with a neural network of neural networks, with [repo](https://github.com/L-Lewis/Predicting-traffic-accidents-CNN) * [Multi-Input Deep Neural Networks with PyTorch-Lightning - Combine Image and Tabular Data](https://rosenfelder.ai/multi-input-neural-network-pytorch/) -> excellent intro article using pytorch, not actually applied to satellite data but to real estate data, with [repo](https://github.com/MarkusRosen/pytorch_multi_input_example) @@ -1318,7 +1368,13 @@ These techniques combine multiple data types, e.g. imagery and text data. * [pytorch-widedeep](https://github.com/jrzaurin/pytorch-widedeep) -> A flexible package for multimodal-deep-learning to combine tabular data with text and images using Wide and Deep models in Pytorch * [accidentRiskMap](https://github.com/songtaohe/accidentRiskMap) -> Inferring high-resolution traffic accident risk maps based on satellite imagery and GPS trajectories -## Few-shot learning +
++ +### Few-shot learning This is a class of techniques which attempt to make predictions for classes with few, one or even zero examples provided during training. In zero shot learning (ZSL) the model is assisted by the provision of auxiliary information which typically consists of descriptions/semantic attributes/word embeddings for both the seen and unseen classes at train time ([ref](https://learnopencv.com/zero-shot-learning-an-introduction/)). These approaches are particularly relevant to remote sensing, where there may be many examples of common classes, but few or even zero examples for other classes of interest. * [Unseen Land Cover Classification from High-Resolution Orthophotos Using Integration of Zero-Shot Learning and Convolutional Neural Networks](https://www.mdpi.com/2072-4292/12/10/1676) * [FSODM](https://github.com/lixiang-ucas/FSODM) -> Official Code for paper "Few-shot Object Detection on Remote Sensing Images" on [arxiv](https://arxiv.org/abs/2006.07826) @@ -1334,7 +1390,13 @@ This is a class of techniques which attempt to make predictions for classes with * [S3Net](https://github.com/ZhaohuiXue/S3Net) -> code for 2022 [paper](https://ieeexplore.ieee.org/document/9791365): S3Net: Spectral–Spatial Siamese Network for Few-Shot Hyperspectral Image Classification * [SiameseNet-for-few-shot-Hyperspectral-Classification](https://github.com/jjwwczy/jjwwczy-SiameseNet-for-few-shot-Hyperspectral-Classification) -> code for 2020 paper: 3DCSN:SiameseNet-for-few-shot-Hyperspectral-Classification -## Self-supervised, unsupervised & contrastive learning +
++ +### Self-supervised, unsupervised & contrastive learning These techniques use unlabelled datasets. [Yann LeCun](https://braindump.jethro.dev/posts/lecun_cake_analogy/) has described self/unsupervised learning as the 'base of the cake': *If we think of our brain as a cake, then the cake base is unsupervised learning. The machine predicts any part of its input for any observed part, all without the use of labelled data. Supervised learning forms the icing on the cake, and reinforcement learning is the cherry on top.* * [Seasonal Contrast: Unsupervised Pre-Training from Uncurated Remote Sensing Data](https://devblog.pytorchlightning.ai/seasonal-contrast-transferable-visual-representations-for-remote-sensing-73a17863ed07) -> Seasonal Contrast (SeCo) is an effective pipeline to leverage unlabeled data for in-domain pre-training of remote sensing representations. Models trained with SeCo achieve better performance than their ImageNet pre-trained counterparts and state-of-the-art self-supervised learning methods on multiple downstream tasks. [paper](https://arxiv.org/abs/2103.16607) and [repo](https://github.com/ElementAI/seasonal-contrast) * [Unsupervised Learning for Land Cover Classification in Satellite Imagery](https://omdena.com/blog/land-cover-classification/) @@ -1381,7 +1443,13 @@ Classification * [SatMAE](https://github.com/sustainlab-group/SatMAE) -> code for 2022 [paper](https://arxiv.org/abs/2207.08051): SatMAE: Pre-training Transformers for Temporal and Multi-Spectral Satellite Imagery * [FireCLR-Wildfires](https://github.com/spaceml-org/FireCLR-Wildfires) -> code for 2022 [paper](https://arxiv.org/abs/2211.14654): Unsupervised Wildfire Change Detection based on Contrastive Learning -## Weakly & semi-supervised learning +
++ +### Weakly & semi-supervised learning These techniques use a partially annotated dataset * [MARE](https://github.com/VMarsocci/MARE) -> self-supervised Multi-Attention REsu-net for semantic segmentation in remote sensing * [SSGF-for-HRRS-scene-classification](https://github.com/weihancug/SSGF-for-HRRS-scene-classification) -> code for 2018 [paper](https://www.sciencedirect.com/science/article/abs/pii/S0924271617303428): A semi-supervised generative framework with deep learning features for high-resolution remote sensing image scene classification @@ -1403,7 +1471,13 @@ These techniques use a partially annotated dataset * [es-CP](https://github.com/majidseydgar/Res-CP) -> code for 2022 [paper](https://ieeexplore.ieee.org/abstract/document/9849704): Semi-Supervised Hyperspectral Image Classification Using a Probabilistic Pseudo-Label Generation Framework * [Flood_Mapping_SSL](https://github.com/YJ-He/Flood_Mapping_SSL) -> code for 2022 [paper](https://ieeexplore.ieee.org/abstract/document/9924583): Enhancement of Urban Floodwater Mapping From Aerial Imagery With Dense Shadows via Semisupervised Learning -## Active learning +
++ +### Active learning Supervised deep learning techniques typically require a huge number of annotated/labelled examples to provide a training dataset. However labelling at scale take significant time, expertise and resources. Active learning techniques aim to reduce the total amount of annotation that needs to be performed by selecting the most useful images to label from a large pool of unlabelled images, thus reducing the time to generate useful training datasets. These processes may be referred to as [Human-in-the-Loop Machine Learning](https://medium.com/pytorch/https-medium-com-robert-munro-active-learning-with-pytorch-2f3ee8ebec) * [Active learning for object detection in high-resolution satellite images](https://arxiv.org/abs/2101.02480) -> arxiv paper * [AIDE V2 - Tools for detecting wildlife in aerial images using active learning](https://github.com/microsoft/aerial_wildlife_detection) @@ -1416,15 +1490,29 @@ Supervised deep learning techniques typically require a huge number of annotated * [Active-Learning-for-Remote-Sensing-Image-Retrieval](https://github.com/flateon/Active-Learning-for-Remote-Sensing-Image-Retrieval) -> unofficial implementation of paper: A Novel Active Learning Method in Relevance Feedback for Content-Based Remote Sensing Image Retrieval * [DIAL](https://github.com/alteia-ai/DIAL) -> code for 2022 [paper](https://arxiv.org/abs/2201.01047): DIAL: Deep Interactive and Active Learning for Semantic Segmentation in Remote Sensing -## Federated learning +
++ + +### Federated learning Federated learning is a process for training models in a distributed fashion without sharing of data * [Federated-Learning-for-Remote-Sensing](https://github.com/anandcu3/Federated-Learning-for-Remote-Sensing) -> implementation of three Federated Learning models -## Image registration +
++ + +### Image registration Image registration is the process of registering one or more images onto another (typically well georeferenced) image. Traditionally this is performed manually by identifying control points (tie-points) in the images, for example using QGIS. This section lists approaches which mostly aim to automate this manual process. There is some overlap with the data fusion section but the distinction I make is that image registration is performed as a prerequisite to downstream processes which will use the registered data as an input. * [Wikipedia article on registration](https://en.wikipedia.org/wiki/Image_registration) -> register for change detection or [image stitching](https://mono.software/2018/03/14/Image-stitching/) * [Phase correlation](https://en.wikipedia.org/wiki/Phase_correlation) is used to estimate the XY translation between two images with sub-pixel accuracy. Can be used for accurate registration of low resolution imagery onto high resolution imagery, or to register a [sub-image on a full image](https://www.mathworks.com/help/images/registering-an-image-using-normalized-cross-correlation.html) -> Unlike many spatial-domain algorithms, the phase correlation method is resilient to noise, occlusions, and other defects. With [additional pre-processing](https://scikit-image.org/docs/dev/auto_examples/registration/plot_register_rotation.html) image rotation and scale changes can also be calculated. - * [How to Co-Register Temporal Stacks of Satellite Images](https://medium.com/sentinel-hub/how-to-co-register-temporal-stacks-of-satellite-images-5167713b3e0b) +* [How to Co-Register Temporal Stacks of Satellite Images](https://medium.com/sentinel-hub/how-to-co-register-temporal-stacks-of-satellite-images-5167713b3e0b) * [ImageRegistration](https://github.com/jandremarais/ImageRegistration) -> Interview assignment for multimodal image registration using SIFT * [imreg_dft](https://github.com/matejak/imreg_dft) -> Image registration using discrete Fourier transform. Given two images it can calculate the difference between scale, rotation and position of imaged features. Used by the [up42 co-registration service](https://up42.com/marketplace/blocks/processing/up42-coregistration) * [arosics](https://danschef.git-pages.gfz-potsdam.de/arosics/doc/about.html) -> Perform automatic subpixel co-registration of two satellite image datasets using phase-correlation, XY translations only. @@ -1461,7 +1549,13 @@ Image registration is the process of registering one or more images onto another * [OSMNet](https://github.com/zhanghan9718/OSMNet) -> code for 2021 [paper](https://ieeexplore.ieee.org/document/9609993): Explore Better Network Framework for High-Resolution Optical and SAR Image Matching * [L2_Siamese](https://github.com/TheKiteFlier/L2_Siamese) -> code for the 2020 [paper](https://ieeexplore.ieee.org/document/9264687): Registration of Multiresolution Remote Sensing Images Based on L2-Siamese Model -## Data fusion +
++ +### Data fusion Data fusion covers techniques which integrate multiple datasources, for example fusing SAR & optical to make predictions about crop type. It can also cover fusion with non imagery data such as IOT sensor data * [Awesome-Data-Fusion-for-Remote-Sensing](https://github.com/px39n/Awesome-Data-Fusion-for-Remote-Sensing) * [UDALN_GRSL](https://github.com/JiaxinLiCAS/UDALN_GRSL) -> Deep Unsupervised Blind Hyperspectral and Multispectral Data Fusion @@ -1500,7 +1594,13 @@ Data fusion covers techniques which integrate multiple datasources, for example * [MANet](https://github.com/caohuimin/MANet) -> code for 2022 [paper](https://www.mdpi.com/2072-4292/14/18/4600): MANet: A Network Architecture for Remote Sensing Spatiotemporal Fusion Based on Multiscale and Attention Mechanisms * [DCSA-Net](https://github.com/Julia90/DCSA-Net) -> code for 2022 [paper](https://www.mdpi.com/2072-4292/14/19/4941): Dynamic Convolution Self-Attention Network for Land-Cover Classification in VHR Remote-Sensing Images -## Terrain mapping, Disparity Estimation, Lidar, DEMs & NeRF +
++ +### Terrain mapping, Disparity Estimation, Lidar, DEMs & NeRF Measure surface contours & locate 3D points in space from 2D images. NeRF stands for Neural Radiance Fields and is the term used in deep learning communities to describe a model that generates views of complex 3D scenes based on a partial set of 2D images * [Wikipedia DEM article](https://en.wikipedia.org/wiki/Digital_elevation_model) and [phase correlation](https://en.wikipedia.org/wiki/Phase_correlation) article * [Intro to depth from stereo](https://github.com/IntelRealSense/librealsense/blob/master/doc/depth-from-stereo.md) @@ -1542,13 +1642,25 @@ Measure surface contours & locate 3D points in space from 2D images. NeRF stands * [GF-7_Stereo_Matching](https://github.com/Sheng029/GF-7_Stereo_Matching) -> code for paper: Large Scene DSM Generation of Gaofen-7 Imagery Combined with Deep Learning * [Mapping drainage ditches in forested landscapes using deep learning and aerial laser scanning](https://github.com/williamlidberg/Mapping-drainage-ditches-in-forested-landscapes-using-deep-learning-and-aerial-laser-scanning) -## Thermal Infrared +
++ +### Thermal Infrared * [The World Needs (a lot) More Thermal Infrared Data from Space](https://towardsdatascience.com/the-world-needs-a-lot-more-thermal-infrared-data-from-space-dbbba389be8a) * [IR2VI thermal-to-visible image translation framework based on GANs](https://arxiv.org/abs/1806.09565) with [code](https://github.com/linty5/IR2VI_CycleGAN) * [Object_Classification_in_Thermal_Images](https://www.researchgate.net/publication/328400392_Object_Classification_in_Thermal_Images_using_Convolutional_Neural_Networks_for_Search_and_Rescue_Missions_with_Unmanned_Aerial_Systems) -> classification accuracy was improved by adding the object size as a feature directly within the CNN * [Thermal imaging with satellites](https://chrieke.medium.com/thermal-imaging-with-satellites-34f381856dd1) blog post by Christoph Rieke -# SAR +
++ +### SAR * [awesome-sar](https://github.com/RadarCODE/awesome-sar) -> A curated list of awesome Synthetic Aperture Radar (SAR) software, libraries, and resources * [Removing speckle noise from Sentinel-1 SAR using a CNN](https://medium.com/upstream/denoising-sentinel-1-radar-images-5f764faffb3e) * [You do not need clean images for SAR despeckling with deep learning](https://towardsdatascience.com/you-do-not-need-clean-images-for-sar-despeckling-with-deep-learning-fe9c44350b69) -> How Speckle2Void learned to stop worrying and love the noise @@ -1595,7 +1707,13 @@ Measure surface contours & locate 3D points in space from 2D images. NeRF stands * [cogsima2022](https://github.com/galatolofederico/cogsima2022) -> code for 2022 [paper](https://ieeexplore.ieee.org/document/9830661): Enhancing land subsidence awareness via InSAR data and Deep Transformers * [XAI4SAR-PGIL](https://github.com/Alien9427/XAI4SAR-PGIL) -> code for 2021 [paper](https://arxiv.org/abs/2110.14144): Physically Explainable CNN for SAR Image Classification -## NDVI - vegetation index +
++ +### NDVI - vegetation index * Calculated via band math `ndvi = np.true_divide((ir - r), (ir + r))` but challenging due to the size of the imagery * [Example notebook local](http://nbviewer.jupyter.org/github/HyperionAnalytics/PyDataNYC2014/blob/master/ndvi_calculation.ipynb) * [Landsat data in cloud optimised (COG) format analysed for NDVI](https://github.com/pangeo-data/pangeo-example-notebooks/blob/master/landsat8-cog-ndvi.ipynb) with [medium article here](https://medium.com/pangeo/cloud-native-geoprocessing-of-earth-observation-satellite-data-with-pangeo-997692d91ca2). @@ -1606,7 +1724,13 @@ Measure surface contours & locate 3D points in space from 2D images. NeRF stands * [Awesome-Vegetation-Index](https://github.com/px39n/Awesome-Vegetation-Index) * [Remote-Sensing-Indices-Derivation-Tool](https://github.com/rander38/Remote-Sensing-Indices-Derivation-Tool) -> Calculate spectral remote sensing indices from satellite imagery -## General image quality +
++ +### General image quality * Convolutional autoencoder network can be employed to image denoising, [read about this on the Keras blog](https://blog.keras.io/building-autoencoders-in-keras.html) * [jitter-compensation](https://github.com/caiya55/jitter-compensation) -> Remote Sensing Image Jitter Detection and Compensation Using CNN * [DeblurGANv2](https://github.com/VITA-Group/DeblurGANv2) -> Deblurring (Orders-of-Magnitude) Faster and Better @@ -1624,7 +1748,13 @@ Measure surface contours & locate 3D points in space from 2D images. NeRF stands * [deep-hs-prior](https://github.com/acecreamu/deep-hs-prior) -> code for 2019 [paper](https://arxiv.org/abs/1902.00301): Deep Hyperspectral Prior: Denoising, Inpainting, Super-Resolution * [iquaflow](https://github.com/satellogic/iquaflow) -> from Satellogic, an image quality framework that aims at providing a set of tools to assess image quality by using the performance of AI models trained on the images as a proxy. -## Synthetic data +
++ +### Synthetic data Training data can be hard to acquire, particularly for rare events such as change detection after disasters, or imagery of rare classes of objects. In these situations, generating synthetic training data might be the only option. This has become quite sophisticated, with 3D models being use with open source games engines such as [Unreal](https://www.unrealengine.com/en-US/). * [The Synthinel-1 dataset: a collection of high resolution synthetic overhead imagery for building segmentation](https://arxiv.org/ftp/arxiv/papers/2001/2001.05130.pdf) with [repo](https://github.com/timqqt/Synthinel) * [RarePlanes](https://registry.opendata.aws/rareplanes/) -> incorporates both real and synthetically generated satellite imagery including aircraft. Read the [arxiv paper](https://arxiv.org/abs/2006.02963) and checkout [this repo](https://github.com/jdc08161063/RarePlanes). Note the dataset is available through the AWS Open-Data Program for free download @@ -1647,6 +1777,11 @@ Training data can be hard to acquire, particularly for rare events such as chang * [Synthesizing Robustness: Dataset Size Requirements and Geographic Insights](https://avanetten.medium.com/synthesizing-robustness-dataset-size-requirements-and-geographic-insights-a687192e8004) -> Medium article, concludes that synthetic data is most beneficial to the rarest object classes and that extracting utility from synthetic data often takes significant effort and creativity * [rs_img_synth](https://github.com/gbaier/rs_img_synth) -> code for 2020 [paper](https://arxiv.org/abs/2011.11314): Synthesizing Optical and SAR Imagery From Land Cover Maps and Auxiliary Raster Data +
+