ArcGIS Pro, Server and the ArcGIS API for Python all include tools to use AI and Deep Learning to solve geospatial problems, such as feature extraction, pixel classification, and feature categorization. This installer includes a broad collection of components, such as PyTorch, TensorFlow, Fast.ai and scikit-learn, for performing deep learning and machine learning tasks, a total collection of 99 packages. These packages can be used with the Deep Learning Training tools, interactive object detection, by using the arcgis.learn module within the ArcGIS API for Python, and directly imported into your own scripts and tools. Most of the tools in this collection will work on any machine, but common deep learning workflows require a recent NVIDIA graphics processing unit (GPU), and problem sizes are bound by available GPU memory, see the requirements section.
This installer adds all the included packages to the default arcgispro-py3 environment that Pro and Server both ship with, and no additional environments are necessary in order to get started using the tools. If you do create custom environments, these packages will also be included so you can use the same tools in your own custom environments as well.
For an example of the kinds of workflows this installer and ArcGIS enables, see the AI & Deep Learning in the UC 2020 Plenary video
- Deep Learning Libraries Installer for ArcGIS Pro 3.0.3
- Deep Learning Libraries Installer for ArcGIS Pro 3.0—3.0.2
- Deep Learning Libraries Installer for ArcGIS Server 11.0
- Deep Learning Libraries Installer for ArcGIS Server Linux 11.0
Downloads for Previous Releases
- Deep Learning Libraries Installer for ArcGIS Pro 2.9
- Deep Learning Libraries Installer for ArcGIS Pro 2.8
- Deep Learning Libraries Installer for ArcGIS Pro 2.7
- Deep Learning Libraries Installer for ArcGIS Pro 2.6
- Deep Learning Libraries Installer for ArcGIS Server 10.9.1
- Deep Learning Libraries Installer for ArcGIS Server 10.9
- Deep Learning Libraries Installer for ArcGIS Server 10.8.1
- Deep Learning Libraries Installer for ArcGIS Server Linux 10.9.1
- Deep Learning Libraries Installer for ArcGIS Server Linux 10.9
- Deep Learning Libraries Installer for ArcGIS Server Linux 10.8.1
On Windows:
Once you've downloaded the archive for your product, extract the Zip file to a new location, and run the Windows Installer (e.g. ProDeepLearning.msi) on Windows. This will install the deep learning frameworks into the default arcgispro-py3 Python environment, but not any custom environments you've created prior to running this installation. After installation, subsequent clones will also include the full deep learning package set. You'll need to extract the file (not just open the .MSI from within the Zip file) or the installer won't be able to find its contents. After installation, the archive and installer files can be deleted.
On Server Linux:
Extract the .tar.gz archive, e.g. with tar xvf <file>.tar.gz, then run the DeepLearning-Setup.sh script. For Server 10.9 and earlier, this would create a package set inside of the Server runtime environment. Starting at Server 10.9.1, this installation creates a new deeplearning environment located in <Server Install>/framework/runtime/deeplearning and the deep learning packages are all native Linux implementations. Next, please uncomment and update the ARCGIS_CONDA_DEEPLEARNING variable in the <Server Install>/arcgis/server/usr/init_user_param.sh file and restart your ArcGIS Server.
Upgrading From a Previous Version:
If you're upgrading from a previous release, the safest way to upgrade is to uninstall and reinstall both the product and the deep learning installer. For example, to upgrade from Pro 2.8 to Pro 2.9:
- Uninstall Deep Learning Libraries for ArcGIS Pro 2.8
- Uninstall ArcGIS Pro 2.8
- Directly remove any files still present in
C:\Program Files\ArcGIS\Pro\bin\Python\envs\arcgispro-py3or equivalent location for your installation. These may have been left over from previously modified environment. - Install ArcGIS Pro 2.9
- Install ArcGIS Pro 2.9 Deep Learning downloaded from this site.
After these steps, you should have a clean Pro installation with the Deep Learning package set included in the default arcgispro-py3 environment.
Manual Installation:
You can install the libraries manually using these archived instructions:
| Following these steps will install an experimental unverified package set | |
|---|---|
| ℹ️ | Make sure to clone the default Python environment to backup your install (see below) |
You can install the deep learning libraries from a command prompt using these steps:
- Open the
Python Command Promptwindow.- You can search for this command prompt in the
Startmenu on Windows, or you can launch it from the product's install folder. - If running an enterprise product search for the
Python Command Prompt 3
- You can search for this command prompt in the
- Clone the default Python environment with this command: (don't forget the
--pinned!)`conda create -n your-clone-name --clone arcgispro-py3 --pinned
- When the Python environment has been cloned, activate the cloned environment:
activate your-clone-name- When the cloned enviornment is activated, the new environment name appears at the beginning of the path:
(your-clone-name) C:\Program Files\ArcGIS\Pro\bin\Python\envs> - Install the deep learning essentials libraries into your cloned environment with:
conda install deep-learning-essentials- When prompted to proceed, review the information, type
y, and pressEnter - If the packages install successfully your cloned enviornment is now setup to run deep learning workflows
- When prompted to proceed, review the information, type
- Type the following command to swap your product's default enviornment to your new cloned environment:
proswap your-clone-name- When you next launch your product it will launch with
your-clone-nameas the active Python Environment and you should now be able to use deep learning tools
- When you next launch your product it will launch with
- If you run into any issues please contact Esri Technical Support
If you will be working in a disconnected environment, download the arcgis_dl_backbones package from the links below and follow the instructions under the Steps to Install listed on the package page. The package places backbones for deep learning models in the specified install location, eliminating the need for internet access when training deep learning models in ArcGIS.
| Pro versions | Server versions | Backbones package |
|---|---|---|
| 2.9—3.0 | 10.9.1—11.0 | PyTorch 1.8 arcgis_dl_backbones package |
| 2.5—2.8 | 10.7.1—10.9.0 | PyTorch 1.4 arcgis_dl_backbones package |
Once you've installed the deep learning libraries, you can use the Deep Learning Tools to train geospatial deep learning models. You can also find out more about the capabilities of the arcgis.learn module which provides specialized access to many geospatial models beyond those directly available as Geoprocessing tools. Finally, you can add any of the above libraries to your own workflows, by importing the packages listed below.
- Learn ArcGIS lesson on using Deep Learning to Access Palm Tree Health
- Join us later in 2021 for the Spatial Data Science MOOC
A collection of recent User Conference 2020 Technical Workshops on Deep Learning:
- Deep Dive into Deep Learning
- Deep Learning for Geographers
- Using Deep Learning with Imagery in ArcGIS
Most of the packages included in the Deep Learning Libraries installer will work out of the box on any machine configuration. For example, PyTorch optionally can take advantage of a GPU, but will fall back to running its calculations on the CPU if a GPU is not available. However, GPU computation is significantly faster, and some packages such as TensorFlow in this distribution only will work with a supported GPU. CUDA, or Compute Unified Device Architecture, is a general purpose computing platform for GPUs, a requirement for current GPU backed deep learning tools.
| GPU requirement | Supported |
|---|---|
| GPU Type | NVIDIA with CUDA Compute Capability 3.7 minimum, 6.1 or later recommended. See the list of CUDA-enabled cards to determine the compute capability of a GPU. |
| GPU driver | NVIDIA GPU drivers — version 456.38 or higher is required. |
| Dedicated graphics memory † | minimum: 4GB recommended: 8GB or more, depending on the deep learning model architecture and the batch size being used |
† GPU memory, unlike system memory, cannot be accessed 'virtually'. If a model training consumes more GPU memory than you have available, it will fail. GPU memory is also shared across all uses of the machine, so open Pro projects with maps and other applications can limit the available memory for use with these tools.
ℹ️ An out-of-date GPU driver will cause deep learning tools to fail with runtime errors indicating that CUDA is not installed or an unsupported toolchain is present. Verify that you have up-to-date GPU drivers directly provided by NVIDIA.
| Library Name | Version | Description |
|---|---|---|
| absl-py | 0.13.0 | Abseil Python Common Libraries |
| addict | 2.4.0 | Provides a dictionary whose items can be set using both attribute and item syntax |
| aiohttp | 3.7.4.post0 | Async http client/server framework (asyncio) |
| alembic | 1.6.4 | A database migration tool for SQLAlchemy |
| astor | 0.8.1 | Read, rewrite, and write Python ASTs nicely |
| astunparse | 1.6.3 | An AST unparser for Python |
| async-timeout | 3.0.1 | Timeout context manager for asyncio programs |
| beautifulsoup4 | 4.10.0 | Python library designed for screen-scraping |
| boost | 1.73.0 | Boost provides peer-reviewed portable C++ source libraries |
| bottleneck | 1.3.2 | Fast NumPy array functions written in Cython |
| catalogue | 1.0.0 | Super lightweight function registries for your library |
| catboost | 0.26 | Gradient boosting on decision trees library |
| category_encoders | 2.2.2 | A collection sklearn transformers to encode categorical variables as numeric |
| charset-normalizer | 2.0.4 | A fast and robust universal character set detector |
| cliff | 3.8.0 | Command Line Interface Formulation Framework |
| cloudpickle | 2.0.0 | Extended pickling support for Python objects |
| cmaes | 0.8.2 | Blackbox optimization with the Covariance Matrix Adaptation Evolution Strategy |
| cmd2 | 2.1.1 | A tool for building interactive command line apps |
| colorlog | 5.0.1 | Log formatting with colors! |
| colour | 0.1.5 | Python color representations manipulation library (RGB, HSL, web, ...) |
| coverage | 5.5 | Code coverage measurement for Python |
| cudatoolkit | 11.1.1 | NVIDIA's CUDA toolkit |
| cudnn | 8.1.0.77 | NVIDIA's cuDNN deep neural network acceleration library |
| cymem | 2.0.5 | Manage calls to calloc/free through Cython |
| cython | 0.29.24 | The Cython compiler for writing C extensions for the Python language |
| cython-blis | 0.4.1 | Fast matrix-multiplication as a self-contained Python library – no system dependencies! |
| cytoolz | 0.11.0 | Cython implementation of Toolz. High performance functional utilities |
| dask-core | 2021.10.0 | Parallel Python with task scheduling |
| dataclasses | 0.8 | A backport of the dataclasses module for Python 3.6 |
| deep-learning-essentials | 2.9 | Expansive collection of deep learning packages |
| dtreeviz | 1.3 | Decision tree visualization |
| fastai | 1.0.63 | fastai makes deep learning with PyTorch faster, more accurate, and easier |
| fastprogress | 0.2.3 | A fast and simple progress bar for Jupyter Notebook and console |
| fasttext | 0.9.2 | Efficient text classification and representation learning |
| filelock | 3.3.1 | A platform independent file lock |
| fsspec | 2021.8.1 | A specification for pythonic filesystems |
| gast | 0.3.3 | Python AST that abstracts the underlying Python version |
| geos | 3.5.0 | A C++ port of the Java Topology Suite (JTS) |
| google-auth | 1.33.0 | Google authentication library for Python |
| google-auth-oauthlib | 0.4.1 | Google Authentication Library, oauthlib integration with google-auth |
| google-pasta | 0.2.0 | pasta is an AST-based Python refactoring library |
| googledrivedownloader | 0.4 | Minimal class to download shared files from Google Drive |
| graphviz | 2.38 | Open Source graph visualization software |
| grpcio | 1.36.1 | HTTP/2-based RPC framework |
| imageio | 2.8.0 | A Python library for reading and writing image data |
| imgaug | 0.4.0 | Image augmentation for machine learning experiments |
| joblib | 1.1.0 | Python function as pipeline jobs |
| keepalive | 0.5 | urllib keepalive support for Python |
| keras-gpu | 2.4.3 | Deep Learning Library for Theano and TensorFlow |
| keras-preprocessing | 1.1.2 | Data preprocessing and data augmentation module of the Keras deep learning library |
| laspy | 1.7.0 | A Python library for reading, modifying and creating LAS files |
| libboost | 1.73.0 | Free peer-reviewed portable C++ source libraries |
| libopencv | 4.5.2 | Computer vision and machine learning software library |
| libuv | 1.40.0 | Cross-platform asynchronous I/O |
| libwebp | 1.2.0 | WebP image library |
| libxgboost | 1.3.3 | eXtreme Gradient Boosting |
| lightgbm | 3.2.1 | LightGBM is a gradient boosting framework that uses tree based learning algorithms |
| llvmlite | 0.37.0 | A lightweight LLVM python binding for writing JIT compilers |
| lmdb | 0.9.29 | Universal Python binding for the LMDB 'Lightning' Database |
| locket | 0.2.1 | File-based locks for Python for Linux and Windows |
| mako | 1.1.4 | Template library written in Python |
| markdown | 3.3.4 | Python implementation of Markdown |
| mljar-supervised | 0.10.6 | Automated Machine Learning Pipeline with Feature Engineering and Hyper-Parameters Tuning |
| mmcv-full | 1.3.7 | OpenMMLab Computer Vision Foundation |
| mmdet | 2.13.0 | OpenMMLab Computer Vision Foundation |
| mmsegmentation | 0.14.1 | semantic segmentation toolbox and benchmark |
| multidict | 5.1.0 | Key-value pairs where keys are sorted and can reoccur |
| murmurhash | 1.0.5 | A non-cryptographic hash function |
| nb_conda_kernels | 2.3.1 | Launch Jupyter kernels for any installed conda environment |
| ninja | 1.10.2 | A small build system with a focus on speed |
| numba | 0.54.1 | NumPy aware dynamic Python compiler using LLVM |
| nvidia-ml-py3 | 7.352.0 | Python bindings to the NVIDIA Management Library |
| onnx | 1.9.0 | Open Neural Network Exchange library |
| onnx-tf | 1.8.0 | Experimental Tensorflow Backend for ONNX |
| opencv | 4.5.2 | Computer vision and machine learning software library |
| optuna | 2.8.0 | A hyperparameter optimization framework |
| opt_einsum | 3.3.0 | Optimizing einsum functions in NumPy, Tensorflow, Dask, and more with contraction order optimization |
| partd | 1.2.0 | Data structure for on-disk shuffle operations |
| patsy | 0.5.2 | Describing statistical models in Python using symbolic formulas |
| pbr | 5.6.0 | Python Build Reasonableness |
| plac | 1.1.0 | The smartest command line arguments parser in the world |
| plotly | 4.5.4 | An interactive, browser-based graphing library for Python |
| pooch | 1.0.0 | A friend to fetch your Python library's sample data files |
| preshed | 3.0.2 | Cython Hash Table for Pre-Hashed Keys |
| prettytable | 2.1.0 | Display tabular data in a visually appealing ASCII table format |
| py-boost | 1.73.0 | Free peer-reviewed portable C++ source libraries |
| py-opencv | 4.5.2 | Computer vision and machine learning software library |
| py-xgboost | 1.3.3 | Python bindings for the scalable, portable and distributed gradient boosting XGBoost library |
| py4j | 0.10.9.2 | Enables Python programs to dynamically access arbitrary Java objects |
| pyasn1 | 0.4.8 | ASN.1 types and codecs |
| pyasn1-modules | 0.2.8 | A collection of ASN.1-based protocols modules |
| pyclipper | 1.3.0 | Cython wrapper of Angus Johnson's Clipper library for polygon clipping |
| pycocotools | 2.0.2 | Python API for the MS-COCO dataset |
| pyperclip | 1.8.2 | A cross-platform clipboard module for Python |
| pyreadline | 2.1 | A python implmementation of GNU readline |
| pyspark | 3.1.2 | Apache Spark |
| python-editor | 1.0.4 | Programmatically open an editor, capture the result |
| python-flatbuffers | 2.0 | Python runtime library for use with the Flatbuffers serialization format |
| python-graphviz | 0.16 | Simple Python interface for Graphviz |
| python-levenshtein | 0.12.2 | Python extension for computing string edit distances and similarities |
| python-lmdb | 1.2.1 | Universal Python binding for the LMDB 'Lightning' Database |
| python_abi | 3.7 | Metapackage to select Python implementation |
| pytorch | 1.8.2 | PyTorch is an optimized tensor library for deep learning using GPUs and CPUs |
| pywavelets | 1.1.1 | Discrete Wavelet Transforms in Python |
| rdflib | 5.0.0 | RDFLib is a pure Python package for working with RDF |
| retrying | 1.3.3 | Simplify the task of adding retry behavior to just about anything |
| rsa | 4.7.2 | Pure-Python RSA implementation |
| sacremoses | 0.0.43 | SacreMoses |
| scikit-image | 0.17.2 | Image processing routines for SciPy |
| scikit-learn | 1.0.1 | A set of python modules for machine learning and data mining |
| scikit-plot | 0.3.7 | Plotting for scikit-learn objects |
| seaborn | 0.11.2 | Statistical data visualization |
| sentencepiece | 0.1.91 | Unsupervised text tokenizer and detokenizer |
| shap | 0.39.0 | A unified approach to explain the output of any machine learning model |
| shapely | 1.7.0 | Geometric objects, predicates, and operations |
| slicer | 0.0.7 | A small package for big slicing |
| soupsieve | 2.2.1 | A modern CSS selector implementation for BeautifulSoup |
| spacy | 2.2.4 | Industrial-strength Natural Language Processing |
| sparqlwrapper | 1.8.5 | SPARQL Endpoint interface to Python for use with rdflib |
| srsly | 1.0.2 | Modern high-performance serialization utilities for Python |
| statsmodels | 0.12.2 | Statistical computations and models |
| stevedore | 3.3.0 | Manage dynamic plugins for Python applications |
| tabulate | 0.8.9 | Pretty-print tabular data in Python, a library and a command-line utility |
| tbb | 2021.4.0 | High level abstract threading library |
| tensorboard | 2.6.0 | TensorBoard lets you watch Tensors Flow |
| tensorboard-data-server | 0.6.0 | Data server for TensorBoard |
| tensorboard-plugin-wit | 1.6.0 | What-If Tool TensorBoard plugin |
| tensorboardx | 2.2 | TensorBoardX lets you watch Tensors Flow without Tensorflow |
| tensorflow-addons | 0.13.0 | Useful extra functionality for TensorFlow |
| tensorflow-estimator | 2.5.0 | TensorFlow Estimator |
| tensorflow-gpu | 2.5.1 | Metapackage for selecting the GPU-backed TensorFlow variant |
| termcolor | 1.1.0 | ANSII Color formatting for output in terminal |
| terminaltables | 3.1.0 | Generate simple tables in terminals from a nested list of strings |
| thinc | 7.4.0 | Learn super-sparse multi-class models |
| threadpoolctl | 2.2.0 | Python helpers to control the threadpools of native libraries |
| tifffile | 2020.10.1 | Read and write TIFF files |
| tokenizers | 0.10.1 | Fast State-of-the-Art Tokenizers optimized for Research and Production |
| toolz | 0.11.1 | A functional standard library for Python |
| torch-cluster | 1.5.9 | Extension library of highly optimized graph cluster algorithms for use in PyTorch |
| torch-geometric | 1.7.2 | Geometric deep learning extension library for PyTorch |
| torch-scatter | 2.0.7 | Extension library of highly optimized sparse update (scatter and segment) operations |
| torch-sparse | 0.6.10 | Extension library of optimized sparse matrix operations with autograd support |
| torch-spline-conv | 1.2.1 | PyTorch implementation of the spline-based convolution operator of SplineCNN |
| torchvision | 0.9.2 | Image and video datasets and models for torch deep learning |
| tqdm | 4.62.3 | A Fast, Extensible Progress Meter |
| transformers | 4.5.1 | State-of-the-art Natural Language Processing for TensorFlow 2.0 and PyTorch |
| typeguard | 2.12.1 | Runtime type checker for Python |
| typing-extensions | 3.10.0.2 | Backported and Experimental Type Hints for Python |
| wasabi | 0.6.0 | A lightweight console printing and formatting toolkit |
| werkzeug | 2.0.2 | The Python WSGI Utility Library |
| wordcloud | 1.8.1 | A little word cloud generator in Python |
| xgboost | 1.3.3 | Scalable, portable and distributed Gradient Boosting (GBDT, GBRT or GBM) library |
| yapf | 0.31.0 | A formatter for Python files |
| yarl | 1.6.3 | Yet another URL library |
Manifest for Pro 2.8 / Server 10.9.0
| Library Name | Version | Description |
|---|---|---|
| absl-py | 0.12.0 | Abseil Python Common Libraries, see https://github.com/abseil/abseil-py. |
| ase | 3.19.1 | Set of tools for atomistic simulations |
| astor | 0.8.1 | Read, rewrite, and write Python ASTs nicely |
| beautifulsoup4 | 4.9.3 | Python library designed for screen-scraping |
| boost | 1.73.0 | Free peer-reviewed portable C++ source libraries. |
| cachetools | 4.2.2 | Extensible memoizing collections and decorators |
| catalogue | 1.0.0 | Super lightweight function registries for your library |
| cloudpickle | 1.6.0 | Extended pickling support for Python objects |
| cudatoolkit | 10.1.243 | NVIDIA CUDA toolkit |
| cudnn | 7.6.5 | NVIDIA's cuDNN deep neural network acceleration library |
| cymem | 2.0.5 | Manage calls to calloc/free through Cython |
| cython | 0.29.23 | The Cython compiler for writing C extensions for the Python language |
| cython-blis | 0.4.1 | Fast matrix-multiplication as a self-contained Python library – no system dependencies! |
| cytoolz | 0.11.0 | Cython implementation of Toolz. High performance functional utilities |
| dask-core | 2021.5.0 | Parallel Python with task scheduling |
| deep-learning-essentials | 2.8 | A collection of the essential packages to work with deep learning packages and ArcGIS Pro. |
| fastai | 1.0.60 | fastai makes deep learning with PyTorch faster, more accurate, and easier |
| fastprogress | 0.2.3 | A fast and simple progress bar for Jupyter Notebook and console. |
| fasttext | 0.9.2 | fastText - Library for efficient text classification and representation learning |
| filelock | 3.0.12 | A platform independent file lock. |
| fsspec | 0.9.0 | A specification for pythonic filesystems |
| gast | 0.2.2 | Python AST that abstracts the underlying Python version |
| google-auth | 1.21.3 | Google authentication library for Python |
| google-auth-oauthlib | 0.4.2 | Google Authentication Library, oauthlib integration with google-auth |
| google-pasta | 0.2.0 | pasta is an AST-based Python refactoring library |
| googledrivedownloader | 0.4 | Minimal class to download shared files from Google Drive. |
| graphviz | 2.38 | Open Source graph visualization software. |
| grpcio | 1.35.0 | HTTP/2-based RPC framework |
| imageio | 2.8.0 | A Python library for reading and writing image data |
| joblib | 1.0.1 | Lightweight pipelining: using Python functions as pipeline jobs. |
| keepalive | 0.5 | An HTTP handler for urllib that supports HTTP 1.1 and keepalive |
| keras-applications | 1.0.8 | Applications module of the Keras deep learning library. |
| keras-gpu | 2.3.1 | Deep Learning Library for Theano and TensorFlow |
| keras-preprocessing | 1.1.2 | Data preprocessing and data augmentation module of the Keras deep learning library |
| laspy | 1.7.0 | A Python library for reading, modifying and creating LAS files |
| libboost | 1.73.0 | Free peer-reviewed portable C++ source libraries |
| libopencv | 4.5.0 | Computer vision and machine learning software library. |
| libprotobuf | 3.14.0 | Protocol Buffers - Google's data interchange format. C++ Libraries and protoc, the protobuf compiler. |
| libwebp | 1.2.0 | WebP image library |
| llvmlite | 0.36.0 | A lightweight LLVM python binding for writing JIT compilers. |
| locket | 0.2.1 | File-based locks for Python for Linux and Windows |
| markdown | 3.3.4 | Python implementation of Markdown. |
| murmurhash | 1.0.5 | Cython bindings for MurmurHash2 |
| ninja | 1.10.2 | A small build system with a focus on speed |
| numba | 0.53.0 | NumPy aware dynamic Python compiler using LLVM |
| nvidia-ml-py3 | 7.352.0 | Python bindings to the NVIDIA Management Library |
| onnx | 1.7.0 | Open Neural Network Exchange library |
| onnx-tf | 1.5.0 | Experimental Tensorflow Backend for ONNX |
| opencv | 4.5.0 | Computer vision and machine learning software library. |
| opt_einsum | 3.3.0 | Optimizing einsum functions in NumPy, Tensorflow, Dask, and more with contraction order optimization. |
| partd | 1.2.0 | Data structure for on-disk shuffle operations |
| plac | 1.1.0 | The smartest command line arguments parser in the world |
| plotly | 4.5.4 | An interactive JavaScript-based visualization library for Python |
| pooch | 1.0.0 | A friend to fetch your Python library's sample data files |
| preshed | 3.0.2 | Cython Hash Table for Pre-Hashed Keys |
| protobuf | 3.14.0 | Protocol Buffers - Google's data interchange format. |
| py-boost | 1.73.0 | Free peer-reviewed portable C++ source libraries. |
| py-opencv | 4.5.0 | Computer vision and machine learning software library. |
| pyasn1 | 0.4.8 | ASN.1 types and codecs |
| pyasn1-modules | 0.2.8 | A collection of ASN.1-based protocols modules. |
| pytorch | 1.4.0 | PyTorch is an optimized tensor library for deep learning using GPUs and CPUs. |
| pywavelets | 1.1.1 | Discrete Wavelet Transforms in Python |
| rdflib | 5.0.0 | Library for working with RDF, a simple yet powerful language for representing information. |
| retrying | 1.3.3 | Simplify the task of adding retry behavior to just about anything. |
| rsa | 4.7.2 | Pure-Python RSA implementation |
| sacremoses | 0.0.43 | Python port of Moses tokenizer, truecaser and normalizer. |
| scikit-image | 0.17.2 | Image processing routines for SciPy |
| scikit-learn | 0.23.2 | A set of python modules for machine learning and data mining |
| sentencepiece | 0.1.91 | SentencePiece python wrapper |
| soupsieve | 2.2.1 | A modern CSS selector implementation for BeautifulSoup |
| spacy | 2.2.4 | Industrial-strength Natural Language Processing |
| sparqlwrapper | 1.8.5 | SPARQL Endpoint interface to Python for use with rdflib |
| srsly | 1.0.2 | Modern high-performance serialization utilities for Python |
| tensorboard | 2.4.0 | TensorBoard lets you watch Tensors Flow |
| tensorboard-plugin-wit | 1.6.0 | What-If Tool TensorBoard plugin |
| tensorboardx | 2.1 | Tensorboard for PyTorch. |
| tensorflow | 2.1.0 | TensorFlow is a machine learning library. |
| tensorflow-addons | 0.9.1 | Useful extra functionality for TensorFlow 2.x |
| tensorflow-base | 2.1.0 | Base GPU package, tensorflow only. |
| tensorflow-estimator | 2.1.0 | TensorFlow Estimator |
| tensorflow-gpu | 2.1.0 | Metapackage for selecting a TensorFlow variant. |
| termcolor | 1.1.0 | ANSII Color formatting for output in terminal. |
| thinc | 7.4.0 | Learn super-sparse multi-class models |
| threadpoolctl | 2.1.0 | Python helpers to control the threadpools of native libraries |
| tifffile | 2020.10.1 | Read and write image data from and to TIFF files. |
| tokenizers | 0.8.1 | Fast State-of-the-Art Tokenizers optimized for Research and Production |
| toolz | 0.11.1 | A functional standard library for Python |
| torch-cluster | 1.5.4 | Extension library of highly optimized graph cluster algorithms for use in PyTorch |
| torch-geometric-1.5.0 | Geometric deep learning extension library for PyTorch | |
| torch-scatter | 2.0.4 | Extension library of highly optimized sparse update (scatter and segment) operations |
| torch-sparse | 0.6.1 | Extension library of optimized sparse matrix operations with autograd support |
| torch-spline-conv | 1.2.0 | PyTorch implementation of the spline-based convolution operator of SplineCNN |
| torchvision | 0.5.0 | image and video datasets and models for torch deep learning |
| tqdm | 4.59.0 | A Fast, Extensible Progress Meter |
| transformers | 3.3.0 | State-of-the-art Natural Language Processing for TensorFlow 2.0 and PyTorch |
| typeguard | 2.7.0 | Runtime type checker for Python |
| wasabi | 0.6.0 | A lightweight console printing and formatting toolkit |
| werkzeug | 0.16.1 | The comprehensive WSGI web application library. |
- Though this package distributes the GPU based versions of packages, CPU versions can still be installed and used on any machine Pro supports. To install TensorFlow for the CPU, from the Python backstage you can install the
tensorflow-mklpackage to get a CPU only version. - This installer adds packages to the default
arcgispro-py3environment. Any subsequent clones of that environment will also include this full collection of packages. This collection of packages is validated and tested against the version of Pro is installed alongside, and upgrades of Pro will also require reinstallation of the deep learning libraries. Note that when you upgrade the software to a new release, you'll need to uninstall the Deep Learning Libraries installation as well as Pro or Server, and reinstall the new version of this package for that release. - This installer is only available for ArcGIS Pro 2.6+, and ArcGIS Server 10.8.1+ -- for earlier releases, you'll need to follow the documentation for that release on installing the packages through the Python backstage or Python command prompt.
- If you want these packages for a specific environment only, you can install the
deep-learning-essentialspackage which has the same list of dependencies as a standalone conda metapackage.