Robovision AI 5.2 release notes¶

Release date: August 4, 2023

What's new¶

Robovision AI 5.2 is now compatible with Edge. This allows you to exchange your data and models between the platform and the edge device.

You can now create projects of a multilabel classification type. A multilabel classification model helps classify data by predicting any number of labels for a single image.

The project details page now gives you an overview of all the elements of your project. For more information, see Project details.

In the label center, users can now set up the confusion matrix to compare the labels of two labelers. For more information, see View confusion matrix.

You can now see what version of an algorithm was used to create a project. The version is displayed in the project list, project details, and filters.
If you try opening the project whose algorithm is not installed on the deployment, you will get a corresponding message saying that the algorithm is missing.

Filtering samples with the charts used to produce incorrect results. This issue has been fixed.
Before, users could create invalid annotations—annotations outside a sample or zero-surface annotations. These could cause model training to fail. This issue has been fixed.
Now, if you update the annotations while the dataset setup is open, the number of samples in the dataset will get automatically updated.
You can now create classes with both lower- and upper-case letters.

We have fixed the issue where training could fail due to disk cache being out of space.
The training metrics on the model comparison page used to be limited to 1,000 datapoints. Now, the graphs display all available datapoints.
Models with failed or stopped training can now be used to predict annotations, transfer learning, test, and run inference.
We have improved the color difference on the model comparison page. Now, every training is distinctly visualized in the training metrics.

The SOLOv2 prediction performance when testing models has been significantly improved.
Before, if test results included empty annotations, the samples with those annotations weren’t displayed. Now, if there are samples with empty annotations in the test results, they are displayed.
Tags in the test center are now read-only—the options to edit and delete the tags have been hidden.
Filtering test results with the class distribution used to produce incorrect results. This issue has been fixed.

The Robovision AI UI has been optimized for the browser size of 1920x1080.
You will be logged out of the platform if your data import has been in the Uploading stage for more than an hour.
If the storage is full, some functions may become unavailable, for example, saving annotation or training a model.

The Robovision AI platform has been designed and validated for Google Chrome 85 and Mozilla Firefox 77 or later.
In Google Chrome, training metrics and a final validation score might not be displayed if an ad blocker is used. To prevent this, mark the Robovision AI URL as trusted in the Google Chrome settings.
In Google Chrome, the thumbnail view can load slowly if you have password managers like Lastpass activated. To prevent this, disable the extension or work in a guest profile window.

Backward compatibility with Robovision AI versions 3.x and 4.x is not foreseen. But upon request, Robovision will provide the needed assistance and conversion scripts.

Types supported: In general and with respect to the model testing functionality, the following image types are supported: JPEG, JPG, PNG, TIFF, BMP, GIF, including a 4^th transparency channel. EXIF orientation metadata is supported.
File names must contain only Latin (ASCII) characters. There are known rendering, packing, and backup/restore issues with file names/object keys that contain non-Latin characters.

2D labeler performance (in particular the Magnetic lasso and Grab cut tools) may degrade for high-resolution images.
You may face the following issues when using the Prediction tool in the label center:
- The prediction process gets hidden if you leave the label center. To view it again, click the Prediction tool.
- The predicted labels are not added to the samples as the prediction process is running.
- The Status section is not updated as the prediction process is running.

During training setup, the Add model button to select a model for transfer learning is enabled even if there is no model available.
Empty annotations are not displayed in the class distribution for training data even if training data contains empty annotations.