Release 1.19

Release 1.19.398(Maintenance release) - Apr 03, 2025

1. Fix for nginx CVE-2025-1974. More details here

2. There are additional base models for anomaly segmentation. All models are accessible through the advanced settings in the anomaly segmentation session submission dialog.
   
   

3. Images uploaded to the image repository in the context of labeling spec, datagen projects, and anomaly segmentation projects were resized to a fixed resolution, resulting in lower-quality results. This is fixed by retaining the original resolution for uploaded images. Also, the maximum size of an image repository is limited to 500 MB. For a larger set of images, please use the Manual Upload dataset feature to bring your images to a Dataset, and then import the images through the catalog.

4. Enhancement to create an Inspection Studio compatible dataset as part of station group registration.

Release 1.19.375(Bugfix release) - Mar 17, 2025

1. Bug fix for an issue where the file path was not updated in full resolution view in an anomaly segmentation session.

Release 1.19.364 - Mar 7, 2025

1. Model Training enhancements

   1. Early stop conditions to stop training if validation accuracy and loss are not improving for a specified number of epochs.

   2. Resume from a snapshot to continue training.

2. Bug fixes in ‘import features into catalog’ functionality.

Release 1.19.340(Bugfix release) - Feb 22, 2025

1. Bug fix for an issue where, for specific videos, the full-resolution image did not match the corresponding thumbnail.

2. Bug fix for model export issue for anomaly segmentation, classification, and object detection models.

Release 1.19.327 - Feb 14, 2025

1. Support exporting an anomaly segmentation model for inference. Only the models created after this upgrade support the export operation.

2. Support exporting a classification and object detection model trained using labels. Only the models created after this upgrade support the export operation.

3. User interface enhancements on Anomaly segmentation, Model training, and Inspection Studio features.

Release 1.19.240 - Dec 24, 2024

Model Co-pilot(Beta)

Training with labeled data

Train classification and object detection models with labeled data with a few clicks. Manage your dataset versions for training, train the model, and evaluate and analyze your model results all in one place. Please go through the following videos for more details.

Anomaly segmentation models

Train an anomaly detection and segmentation model with just tens of good(non-anomalous) image samples. Get going with a model without the need to collect rare defect data and label them before training a model. Watch the videos below for more details.

Inspection Studio(Beta)

Inspection Studio is a suite of capabilities to support end-to-end model deployment, monitoring, and data collection from quality inspection endpoints. This release introduces the feature as a beta capability. Please contact us to learn how to deploy and integrate these capabilities for your use case.

Below is an introductory video on this feature.

Other features

1. Flexible manual sampling of points on plot view by drawing a polygon.
   

2. Manual labeling requests as a way to augment auto-labeling lifecycle. Contact us for more details.

3. Trajectory mode in Video sequence search to search for events where a chosen object follows the trajectory marked in the query images.