Deep Learning with Imagery
Applications of imagery deep learning models in ArcGIS.
Object Classification
Models: FeatureClassifier with ResNet, Inception, VGG backbones
Classifying homes as damaged or not after a forest fire
Object Detection
Models: SingleShotDetector(SSD), RetinaNet, YOLO, FasterRCNN, MMDetection
Detecting swimming pools
Pixel Classification
Models: UNetClassifier, PSPNetClassifier, DeepLab, MMSegmentation
Land cover classification
Instance Segmentation
Models: MaskRCNN
Extracting building footprints
Panoptic Segmentation
Models: MaxDeepLab
Land cover Extraction, Building Footprint and Car Detection
Edge Detection
Models: BDCNEdgeDetector, HEDEdgeDetector, ConnectNet
Mapping residential parcels
Road Extraction
Models: MultiTaskRoadExtractor
Road extraction
Change Detection
Models: STA-Net ChangeDetector
Identifying new construction
Image to Image Translation
Models: Pix2Pix, Pix2PixHD, CycleGAN
Generating optical imagery from SAR data
Image Captioning
Model: ImageCaptioner
Describing geographical features using a caption
Image Enhancement
Models: SuperResolution
Increasing image resolution
Imagery Time Series Classification
Models: PSETAE
Crop Classification