The individual classifications for each entity can now be marked as required in a project's ontology. This will guarantee that the classifications section of the entity's annotation process will be completed by the annotator before submitting the final annotation for review. This feature can be accessed by navigating to a project's ontology in Alegion Control, editing the entity in question, and further editing the classification in question present within the entity's configuration settings. Once in the classification's settings, there should be an optional checkbox marked as "Classification is required". Select this checkbox to require the classification.
Out of Bounds Support
As a follow-up to our release of shared edges, we have added support for making polygon keypoint selections out of bounds of the image. This makes it easier to draw the shapes of polygons that extend to the edge of an image and improves the platform's semantic and panoptic segmentation abilities. Please see the following animation for an example of the feature in action.
This feature allows for two or more polygons to share one or more of their edges. Users can use a number of different methods to generate shared edges including using an existing edge to generate a new tangential polygon or dragging the edges of existing polygons together. The feature additionally scales to be used with multiple polygons allowing for exponential increases in efficiency. The introduction of Shared Edges is a big step forward in improving the Alegion platform's semantic and panoptic segmentation abilities.
Video Annotation v6
We have released our sixth iteration of the video annotation experience for Alegion Control. The primary focus of this revision was to introduce the second version of the video annotation result data format. This new format aims to increase the consistency of the result data with that of image annotation. This will require minor changes to existing customer video annotation workflows but results in a more seamless experience overall. Read more about the specific changes to the format by visiting the resource below.
Image Annotation Overhaul (V3)
A major redesign of the image annotation interface in Alegion Control. The goal of this redesign was to provide a more consistent user experience across both video and image annotation. The new interface takes after the Alegion Control video annotation interface and should be familiar to those who have used it previously.
Enhanced SmartPoly for Image & Video Annotation
A complete rebuild of the SmartPoly annotation feature. This feature allows annotators to quickly select complex polygon shapes in as little as one click. The rebuilt feature possesses significantly improved detection resulting in more accurate annotations and requiring less intervention from the annotator.
Keypoint Connections in Image Annotation
The Keypoint Connections feature released for video annotation last month is now available in image annotation. The functionality is substantially the same so any of the previously released resources on Keypoint Connections will be accurate for image annotation as well. Please see last month's changelog entry for more information.
URL Format for Navigating to Specific Records in Review Portal
As long as your account is approved for review portal use, you may now use static record URLs to bypass the navigation screens and access specific records directly. This allow users to easily review specific record results without having to search for it and without having to navigate through workflow or batches.
Keypoint Connections in Video Annotation
Keypoint Connections builds on the hierarchical entity types introduced in July of last year. The feature allows the annotator to create direct associations between separate entities as well as enables the assignment of child entities to a parent. These entity relationships are visually represented by lines in the annotation preview. This feature is most commonly used to create entity based skeletons but the possibilities are endless.
Direct Upload of Video and Image Assets to Projects
Alegion has a new way for you to directly upload your assets into the Alegion Control platform. This provides a standard method for providing assets to Alegion for customers not making use of one of the supported self-hosting integrations. The primary benefits of this are increased security and ease of use over methods such as email, non-bucket based cloud storage, or physical asset delivery. For customers wanting to utilize one of our supported secure asset self-hosting platforms, please see the documentation on it here. Currently we continue to support AWS and recently added support for GCS. There are plans to add support for Microsoft's Azure product, but it is not currently supported. If using an unsupported self-hosting solution please see this document.
Automatic Validation and Remediation of Video Assets
Up until now, the video validator tool released early last year required users to manually select mitigating solutions to any incompatibilities found with their video assets. Building upon that system, all videos uploaded into the Alegion platform will now automatically be validated for compatibility. Any corrective actions necessary to prepare the video for annotation will be performed without the need for user intervention. This feature will work in conjunction with our upcoming asset upload system to make the process of getting your video assets ready to annotate take as little time as possible.
Google Cloud Storage Integration
Alegion is continuing its efforts to make the process of getting your data into our platform easier. We are working to support as many customer environments as possible to save you time and effort in your annotation efforts. As such, Google Cloud Storage (GCS) will be joining Amazon Web Services (AWS) in our growing list of approved cloud platforms. This enables customers to provide secure access to annotation assets from their Google Cloud Storage environment.
We have enabled the ability to reorder entity classifications in the project ontology for both image and video annotation. While this is mostly a user interface change, it does allow the project creator to prioritize the most frequently utilized classifications by placing them at the top of the list. This is especially useful when working through iterative ontologies on a project which result in additional classifications being added as the ontology is developed.
With video annotation in mind, this feature allows for frame level classifications to be defined in groups with and without interpolation. The primary use case for this feature would be binary statuses that apply to entire frames of the video. Examples of which would be Day/Night, Sun/Rain, Indoors/Outdoors, etc.
Disable Interpolation Option
While linear interpolation for entity positions and interpolation of classifications often yield accurate results and decrease manual annotation time, there are instances where manual placement would be preferable. Therefore we have added a toggle that allows for the interpolation feature to be disabled.
SmartPoly for Image Annotation
SmartPoly allows annotators to quickly select foreground objects by clicking the desired subject. Using a machine learning based segmentation model, a free-form polygon will automatically be drawn around the subject. While some manual refinement may be necessary, this process will significantly reduce the time required for an annotator to manually draw the polygon.
Using a machine learning model, Alegion control is now able to track foreground objects as they move between frames during video annotation. By drawing a bounding box around the desired entity, an annotator can employ the object tracking feature to have the localized bounding box automatically move with the entity. The frequency of the redrawing of the bounding box can be adjusted using a customizable buffer of keyframes. This will increase and decrease the accuracy alongside processing time allowing the annotator to adjust for their specific needs/use case.
Image Annotation in Control
Alegion is continuing to migrate more features to our new SaaS platform, Control. Following video annotation's successful deployment on the platform, image annotation projects are now supported. This will increase consistency for our internal annotators resulting in a higher quality, and quicker turnaround of annotation results for our managed platform customers. Additionally this makes image annotation and its associated features readily available for use among our SaaS customers.
Video Projects in Review Portal
The Alegion Review Portal has been available for image annotation for some time now. However, with the rapidly growing amount of video annotations on the Alegion platform, it is increasingly important to provide our customers with an intuitive method for reviewing video annotation results. Selected users can now use the Alegion Review Portal to approve or deny video annotation results submitted by annotators.
Hierarchical Entities in Control
Parent/child relationships can now be defined for labeled entities in image and video annotation projects. This allows for annotated components to be attributed to a whole to which they belong. While this feature adds plenty of usability on its own, it also lays the groundwork for a number of upcoming features that we look forward to implementing in the coming year.