New
- API-Reference documentation updates across many modules
- Adds hyperlinks to return value objects
- Adds text emphasis to terms within class, property, method and function descriptions
- Adds and expands upon code snippets
- Adds conceptual grouping of classes and submodules in
mapping
module instead of presenting members alphabetically
Guides
- Working with geometries
- Part 3 - Spatial operations on geometries
- Adds
Area and Lengths
sections toPattern using arcgis.geometry.functions: uses server-side geometry engine
section
- Adds
- Deep Learning with ArcGIS
- Deep learning on time series data
- Adds
Sequence Length
section
- Adds
New Guides:
- Adds Using MMSegmentation with
arcgis.learn
- Adds Part 1 - Introduction to the Spatially Enabled DataFrame. Subsequent parts are to be published in upcoming releases.
Samples
GIS analysts and data scientists
- New Samples:
- Wildlife Species Identification in Camera Trap Images
- Multi-class change detection using image segmentation deep learning models
- Detecting deforestation in the Amazon rainforest using unsupervised K-means clustering on satellite imagery
- Coastline classification using Feature Classifier
- Classification of SfM-derived point clouds using deep learning
- Glacial Terminus Extraction using HRNet
- Finetuning Pre-trained Building Footprint Model
Module Changes
arcgis.gis
- Enhances security through added checks to update
certifi
library GIS
- Adds
proxy
keyword argument documentation to handle eitherhttp
orhttps
protocols
- Adds
ContentManager
clone_items()
- Extends support for unhandled
Item
types by allowing for definition, registration and unregistration of custom classes - Adds support for hosted
Map Services
User
- Adds code snippet to
folders
property
- Adds code snippet to
arcgis.gis.server
- Add documentation to clarify return values for
list()
methods onServerManager
,ServicesDirectory
, andServiceManager
arcgis.gis.nb
- New Classes
NotebookDataAccess
- accessed with the
NotebookServer.data_access
property
- accessed with the
NotebookFile
NotebookServer
- Adds
data_access
property
arcgis.gis.workflowmanager
- Adds documentation for:
JobLocation
JobTemplate
Group
SavedSearchesManager
WMRole
arcgis.gis.features
- New Class
EditFeatureJob
- returned when
edit_features(future=True)
- returned when
FeatureLayer
- Adds
future
parameter toedit_features()
so function can run asynchronously
- Adds
- Adds ability to create a Spatially Enabled DataFrame from the Protocol Buffer Format
FeatureSet
- Fixes issue where initializing an object from a file incorrectly formats date fields
arcgis.features.managers
FeatureLayerCollectionManager
- Fixes BUG-000133695 where
overwrite()
converts a hosted Feature Layer to a hosted Table when the source data is Excel
arcgis.raster
arcgis.raster.analytics
- Adds
md_to_upload
parameter to allow publishingImageryLayers
from existing mosaic datasets inArcGIS Online
to: - Adds code snippets for:
create_image_collection()
- Adds documentation to highlight support for local rasters as
input_rasters
argument
- Adds documentation to highlight support for local rasters as
arcgis.raster.functions
- Enhances API Reference documentation with more details and converts parameter lists to tabular format
convolution()
- Adds NumPy array as an option for the
kernel
parameter stretch()
- Adds
colorramp
parameter wndwi()
- Adds documentation
arcgis.geoanalytics
use_proximity
- Adds
group_by_proximity()
tool
arcgis.geometry.functions
- Simplifies how units are specified by introducing two Enumerations. See code snippet for area_and_lengths:
area_and_lengths
:AreaUnits
for input as thearea_unit
parameterLengthUnits
for input as thelength_unit
parameter
arcgis.widgets
- Adds note and code snippet for using
MapView
in a Jupyter Notebook running overhttp
arcgis.learn
- Blocks model training and adds error when training is not supported in ArcGIS Online Notebooks
- Suppresses warning and info messages in ArcGIS Pro when tool execution is not impacted
- Adds support for
torch.distributed
on Windows machines with thegloo
backend - Adds data checks when searching across multiple folders to data to initialize object
- Adds warning message when training classes do not match validation classes for data
Data Preparation Methods
prepare_data()
- Adds warning message when using
prepare_data()
forTabular
models with a batch size less than default because subsequentlr_find()
operations fail - Adds support for multiple image types when
dataset_type=ObjectTracking
- Deprecates unnecessary keyword arguments. See API reference for warning
- Adds warning message when using
prepare_textdata()
- Adds parameter table to reference documentation
Object Classification Models
FeatureClassifier
- Adds parameter table to
show_results()
documentation - Deprecates following methods:
- Adds parameter table to
Object Detection Models
SingleShotDetector
- Fixes
OverflowError
when training that occurs because of error initializing object
- Fixes
Pixel Classification Models
ChangeDetector
- Adds
Classified Tiles
tosupported_datasets
- Adds
Image Translation Models
- New Models
- Fixes
predict()
not working for Multispectral images of more than 4 bands for: Pix2Pix
- Enhances
show_batch()
andshow_results()
by adding gap between ground truth and predictions for similar images - Enhances code for
predict()
andfrom_model()
- Enhances
ImageCaptioner
- Adds support for data exported from
ArcGIS Pro
- Adds support for data exported from
Object Tracking Models
- Adds support for data exported from
ArcGIS Pro
to: DeepSort
- Adds evaluation metrics in addition to
valid_loss
- Adds evaluation metrics in addition to
Feature, Tabular and Timeseries Models
AutoML
- Enhances
from_model()
to handleempty_data
parameter internally - Enhances
fit()
method to save model intemp
directory
- Enhances
Fixes
Guides
Get Started
- Install and set up
- Updates
Install deep learning dependencies
andInstall using Python Command Prompt or Anaconda
section
- Updates
Mapping and visualization
- Smart Mapping
- Updates search string for
USA freeway system
layer
- Updates search string for
Deep Learning with ArcGIS
- Point cloud classification using PointCNN
- Updates terminology to
classification
fromsegmentation
- Adds tips for new features
- Adds 3D tool information
- Refines environment setup steps
- Adds information about deprecated features
- Updates terminology to
- How Multi-Task Road Extractor works ?
- Updates
Implementation in arcgis.learn
andSetting up the environment
sections
- Updates
- Classification of Raw Point Clouds using Deep Learning & generating 3D Building Models
- Updates workflow to reflect enhancements and deprecation in API
- Updates public data to reflect new geoprocessing workflow
- Expands information about parameters and formatting
Module Changes
arcgis.gis
GIS
- Fixes issue with proxy servers that do not support
https
- Fixes issue when initializing with
GIS("Pro")
if not signed into an organization inArcGIS Pro
- Fixes issue when initializing with
GIS("Pro")
if active organization uses Integrated Windows Authentication
- Fixes issue with proxy servers that do not support
Item
arcgis.gis.admin
EnterpriseUsers
- Fixes incorrect request format sent by
get()
method - Fixes BUG-000141402 where
create()
returnsAttributeError
when creatingbuiltin
user
arcgis.gis.server
LogManager
- Fixes BUG-000141231 where
query(export=True, export_type='CSV')
returnsTypeError
arcgis.gis.mission
LogManager
- Fixes BUG-000141231 where
query(export=True, export_type='CSV')
returnsTypeError
arcgis.gis.nb
LogManager
- Fixes
NameError
whenstart_time
andend_time
arguments are used withquery()
- Fixes BUG-000141231 where
query(export=True, export_type='CSV')
returnsTypeError
arcgis.gis.features
- Optimizes request sent to server for default
where
clause ofquery()
method for: FeatureLayer
Table
FeatureLayer
- Fixes
edit_features()
to acceptGeometry.JSON
as input format for geometries - Fixes
AttributeError
whenquery()
used on services that do not cache results GeoAccessor
- the Spatially Enabled DataFrame- Fixes issue with
plot()
and theClass Breaks Renderer
where largest value in the classifiction becomes a class by itself - Fixes issue with
to_featureclass()
changing field names upon export FeatureCollection
- Removes hard-coded image link for default symbology code
arcgis.features.analysis
arcgis.features.use_proximity
- Fixes issue with multiple functions when using dictionaries as input to various parameters
arcgis.features.analyze_patterns
find_point_clusters()
- Fixes ENH-000140250 where
Failed to execute error
returned if theoutput_name
argument has spaces in it
- Fixes ENH-000140250 where
arcgis.features.managers
ParcelFabricManager
- Fixes incorrect property check with
duplicate()
method
arcgis.raster
RasterCollection
ImageryLayer
- Fixes
compute_histograms()
to properly work with time extent passed as a list ofdatetime.date
objects
- Fixes
arcgis.raster.analytics
predict_using_trend_raster
- Fixes incorrect format used in the request sent with a
dimension_values
argument for ArcGIS Online
arcgis.raster.functions
convolution()
- Fixes issue with incorrect output for user-defined kernels
multidimensional_filter()
- Fixes issue where function returns the same data if the input is reclusively aggregated data
compute_change()
- Fixes issue when chaining with other functions
arcgis.geometry
- Fixes BUG-000135965 where
WKT
andWKB
properties ofGeometry
objects were returning non-standardogc
formatting- implements
GeoMet
library to address issue
- implements
- Fixes issue where
JSON
property returned incorrect format
arcgis.geometry.functions
- Fixes missing
distance_unit
parameter and logic todistance()
- Corrects
areas_and_lengths()
documentation
arcgis.mapping
WebMap
- Fixes BUG-000133736 where a
Vector Tile Service
item saved in a WebMap returned "The item you requested cannot be found. The item may have been deleted or you may have entered an incorrect URL" when attempting to open it OfflineMapAreaManager
- Fixes issue with
create()
using theactive_gis
instead of theWebMap
item's gis - Fixes issue where
tile_services
argument is ignored
arcgis.learn
- Fixes extra checkpoint getting saved with
fit(checkpoint='all', early_stopping=True)
on models - Fixes inferencing with multiple
GPUs
on Linux servers - Fixes
OSError: Too many files open
error when callingsave()
on: - Fixes
IndexError
when training of models is interrupted during the first epoch
Data Preparation Methods
prepare_data()
- Fixes incorrect format for
downsample_factor
keyword parameter - Fixes incomplete progress when applying transformations for
dataset_type='ObjectTracking'
- Fixes issue where function exits cleanly even when no valid images are provided
- Fixes incorrect format for
Object Classification Models
FeatureClassifier
- Suppresses unnecessary warning messages and lack of predicted images when visualizing with:
- Fixes
AttributeError
on model intialized withMultiLabel RGB
orMultiSpectral
data and subsequently runningfit(epochs = 10,monitor='multi_label_fbeta')
- Fixes
RuntimeError
when using training data created with images of different sizes
Object Detection Models
MMDetection
- Fixes initialization with
fcos
andvfnet
models with anArcGIS Pro
environment - Suppresses unnecessary warning messages when intializing using data in
cascade_rpn
,fcos
,nas_fcos
, orvfnet
formats
- Fixes initialization with
Pixel Classifiction Models
MMSegmentation
- Fixes issue with
lr_find()
andfit()
failing when a customer configuration is passed as themodel
argument during initialization - Suppresses unncessary warning when initializing with the
dmnet
model
- Fixes issue with
PSPNetClassifier
- Fixes issue with incorrect
accuracy
score written to the emd file
- Fixes issue with incorrect
Object Tracking Models
SiamMask
- Fixes
from_model()
when used with a model output fromsave()
- Fixes
DeepSort
- Fixes output format for
supported_backbones
- Fixes issue where
accuracy
is returning binary values forshow_results()
- Fixes output format for
Feature, Tabular and Timeseries Models
MLModel
- Fixes issue with featureimportances property returning no output
AutoML
- Fixes
from_model()
to handleempty_data
to correctly search when notebook is in a different directory than model - Fixes
predict_proba()
to correctly search field names used as explanatory variables
- Fixes
FullyConnectedNewtork
- Fixes
show_results()
ignoringrows
argument - Fixes
save(publish=True)
properly returning thedlpk
item id - Removes unintended warnings from
predict()
andfeature_importances_()
- Disables
explain
argument onpredict()
which fails with aTypeError
whenTrue
- Fixes