What’s new for GeoAI in the ArcGIS Pro 3.7 Image Analyst extension

Question

Here are the 8 key updates you need to know about GeoAI for Imagery at ArcGIS Pro 3.7:
• New Drawing Tools for Labeling: Single-click point and line tools for fast mask generation.
• Auto-Detect via Features Labeling: Turn legacy point data into precise segmentation boundaries.
• Multiclass AI Labeling: Use text prompts to label multiple object types in one pass.
• Native Video Labeling: Sample and label training data directly from raw video files.
• Expanded Data Review: QA/QC tools now support all types of training data review
• Foundation Model Support: Native integration for remote sensing backbones like DOFA and TerraMind.
• Translate Pixels Tool: Dedicated workflow for image to image translation e.g. super-resolution and image colorization.
• Optimized Inferencing: Up to 4x faster performance for high-volume, folder-based imagery.

Ready to see how these features transform your workflows?

Read the full blog post: https://www.esri.com/arcgis-blog/products/arcgis-pro/geoai/whats-new-for-geoai-in-the-arcgis-pro-3-7-image-analyst-extension

KrishnaDaravath · Answer

Hi Pavan,

Thank you for your response and for answering my questions. This information is very helpful and exciting to learn about. I’ve also sent you a connection request on LinkedIn and look forward to staying connected.

Thank you again.

Sincerely,
Krishna Daravath

PavanYadav · Answer

Hi @KrishnaDaravath

Thanks for reading and for the great feedback! I’m glad you're excited about the new labeling features and performance improvements. To answer your questions:

* Pretrained models: Yes! In addition to supporting these new architectures, we are continuously expanding the ArcGIS Living Atlas with ready-to-use pretrained models. For instance, there are foundational models (such as the Prithvi models) that are packaged as pretrained models ready for deployment.

* Translate Pixels tool: Absolutely. You can use models trained with Pix2Pix, Pix2PixHD, CycleGAN, Super-Resolution, Super-Resolution with SR3, and others trained on your own datasets.

Thanks again for the comment, and let me know how these tools work out for your workflows!

KrishnaDaravath · Answer

These GeoAI updates look really impressive — especially the multiclass AI labeling and native video labeling features. The faster inferencing improvements are exciting too!

* Will the new foundation model support include ready-to-use pretrained models for common imagery workflows?
* Can the Translate Pixels tool be used with custom-trained models as well?

Thanks for sharing these great updates!