This repository provides installers for a comprehensive suite of deep learning and machine learning Python packages, designed to integrate seamlessly with Esri's ArcGIS system. It targets geospatial professionals and developers looking to leverage AI for tasks like feature extraction, pixel classification, and object detection within ArcGIS Pro, Server, and the ArcGIS API for Python. The benefit is a pre-configured, validated environment with over 250 packages, including PyTorch, TensorFlow, and fast.ai, enabling immediate use of advanced analytical capabilities.

How It Works

The project offers installers that add a large collection of Python packages (over 250) to the default ArcGIS Python environment (arcgispro-py3). This approach simplifies setup by avoiding the need for manual environment management or complex dependency resolution. The packages are curated and tested for compatibility with specific ArcGIS versions, ensuring a stable foundation for deep learning workflows. For users needing more control, a deep-learning-essentials metapackage is available for manual installation into custom environments.

Quick Start & Requirements

• Installation: Download the appropriate installer (e.g., .msi for Windows, .tar.gz for Linux Server) for your ArcGIS version and run it. For manual installation into custom environments, use conda create -n your-env --clone arcgispro-py3 --pinned followed by conda activate your-env and conda install deep-learning-essentials.
• Prerequisites: An NVIDIA GPU with CUDA Compute Capability 5.0+ (6.1+ recommended) and NVIDIA GPU drivers (version 527.41 or higher) are required for GPU-accelerated deep learning. Minimum 4GB dedicated GPU memory is recommended, with 8GB+ for larger models.
• Resources: Setup involves downloading and running an installer, typically taking a few minutes. The package collection is extensive, requiring significant disk space.
• Documentation: Esri Deep Learning Frameworks

Highlighted Details

• Includes PyTorch, TensorFlow, Fast.ai, scikit-learn, and numerous specialized libraries (e.g., albumentations, mmcv, transformers).
• Validated package sets for specific ArcGIS Pro and Server versions ensure compatibility.
• Supports both integrated installation into default environments and manual installation for custom environments.
• Provides workarounds for Windows MAX_PATH limitations in older TensorFlow versions.

Maintenance & Community

This repository is maintained by Esri. Links to Esri Technical Support are provided for assistance.

Licensing & Compatibility

The licensing is tied to the ArcGIS software it supports. While the underlying Python packages have their own open-source licenses (e.g., MIT, Apache 2.0), their use is intended within the Esri ecosystem. Commercial use is subject to ArcGIS licensing terms.

Limitations & Caveats

• TensorFlow 2.13 (Pro 3.3) is CPU-only; GPU support for TensorFlow was removed after Pro 3.1. Users requiring GPU-accelerated TensorFlow should use PyTorch or older ArcGIS versions.
• Installation requires specific ArcGIS versions (Pro 2.6+, Server 10.8.1+).
• Upgrading ArcGIS software necessitates uninstalling and reinstalling the deep learning libraries to maintain compatibility.