Version 10.2.2 Update
The new version of eCognition 10.2.2 is now available for download on our webpage. In this release, we enabled opening the project files in the trial version of eCognition, so now it is not only possible to go through the basic tutorials, but also more advanced rule set examples available on our Community webpage. We also fixed a few problems, including the fix for self-intersecting polygon export and performance of the 2D Feature Space Plot. Please check the Release Notes for the details.
Check & Update Your Maintenance Contract
The newest eCognition version is now available for customers with a valid eCognition maintenance contract. For details on the latest features, tools and changes to the software, please refer to the Release Notes.
This maintenance contract will provide you with the latest product enhancements, updates and premier technical support. Customers with a current maintenance contract have already been informed via email including download and licensing information. If you have not received this email, please request a maintenance quote with your Entitlement ID(s) by contacting our sales team.
Deep Learning for Native Point Clouds
A new automatic point cloud classification algorithm utilizes the most cutting edge deep learning techniques and provides users with simple access to high accuracy results via a growing pre-trained model for both terrestrial and aerial sensors. The model is derived directly from point cloud data to assure little to know loss of valuable data characteristics.
Automated Accuracy Assessment
We have modernized our existing Accuracy Assessment Tool to integrate vector-derived validation samples and added techniques to visualize accuracy via a confusion matrix for a single Project or across multiple Projects in a Workspace. In addition, easy-to-use algorithms have been added to automate the calculation of valuable accuracy statistics and confusion matrices.
Increased Native Vector Handling
The team has built out our current native vector tools with a focus on vector linearization to improve the extraction of linear objects. Furthermore, we have boosted the performance of attribute table handling - yes, you can work with attribute tables in eCognition and now even better!
Expanded Integration of Deep Learning Models
Full Google TensorFlow SavedModel integration now enables the use of semantic-, instance segmentation and object detection models. A new instance segmentation algorithm allows users to take advantage of 3rd party SavedModels and directly segment and classify image objects - accelerating the execution of deep learning models and getting you to meaningful results quicker!
The example below demonstrates the results of a 3rd party object detection model applied to Trimble MX9 mobile mapping image data.
