OptimEyes Developer Update Improves ML Model Compatibility
OptimEyes Developer version 25.6.4.1 is now available with a number of updates to make working with TensorFlow ML models easier.
- For ML models that require specific dimensions for input images, the ML Tool can now scale images and ROIs to the required dimensions.
- The ML Tool now displays result row numbers when the box to save row numbers to VES variables is checked. This is useful, for example, with the EfficientNet-B0 model, which returns probabilities of matches to various object categories, but instead of returning the object category numbers, simply returns the probabilities on the row number corresponding to the category.
- The ML Tool now sets a VES variable indicating how many result groups it saved as VES variables. This is useful when iterating through the results within a VisualEyes Script. For example, when using an object detection model, the ML Tool may draw ROI boxes around the objects that were detected. However, since looking up the object categories and attaching object labels to the ROIs requires additional script steps, it is useful to have the number of objects that were detected automatically stored in a variable in order to make processing each object easier.
- The ROI Manager Tool was updated to enable the display of arbitrary labels (such as object category names) with all types of ROIs. Previously, the only text that could be attached directly to an ROI was the ROI name, and only with Point ROIs.
- Additionally, all ROI types can now display their ROI names.
- The text background for text attached to ROIs can now be either transparent or match the color of the ROI.
- The Lookup Tool was updated to fix a bug that could occur when working with files that do not have column labels in their first row.