November 6th, 2024
Changelog (v2.163.0)
New: Introducing SAM2 for Geospatial Images and Object Tracking on Video!
We're thrilled to bring SAM 2 to both Geospatial Images (.tif files) and Video (Beta). With SAM 2, you now have the flexibility to create segmentation masks using two tailored models:
- Rapid: Get results quickly with a faster model, ideal for rapid processing.
- High-Res: For tasks requiring higher precision, this model delivers greater accuracy at a slower pace.
Enhanced Segmentation Experience on Geospatial Images:
Now available on both tiled and non-tiled images, the Interactive Segmentation tools (Point and Bounding Box) allow for precision control from the toolbar. The final segmentation mask adjusts based on your chosen tool and zoom level.
Advanced Segmentation with Smart Tracking on Videos:
In the video interface (Beta), create an initial segmentation mask on any frame using the Interactive Segmentation tools, then activate Smart Tracking to segment your target automatically over the next 50 frames. If SAM2 loses track of the target, it will stop tracking, and the object will no longer appear in subsequent frames. You can pause tracking at any time by pressing ESC
.
Note: Segmentation masks generate larger volumes of data, so for best performance, use tracking on a single target per session. This is a Beta release, and we’re actively working to make the tracking feature more scalable and efficient—stay tuned for updates!
Fix
- [Predictions] Fixed an issue where predictions imported with subjobs did not properly display their subjob tags in the labeling interface when opening an asset.
- [Filters] Resolved an issue where direct searches by labeler name could not be selected from the filters. This functionality is now working as expected.
- [Read-Only mode] Annotation tools are now correctly hidden in the toolbar when in read-only mode, as editing labels is not permitted
- [Autosave label access]Fixed a issue that, in some cases, allowed access to other team members' autosave labels.