Changelog

June 2023 Update

June 12, 2023
A series of cards from V7 Darwin

Auto-Annotate Powered by SAM

Segment Anything Model (SAM) is a powerful instance segmentation model developed by Meta. Integrating it with our Auto-Annotate tool brings about some game-changing improvements. Our enhanced tool now delivers faster, more accurate, and scalable annotations. You can pre-process entire images, creating semantic masks that give you better control over your annotations.

UI Now Uses Accelerated Label Rendering

Accelerated Label Rendering (ALR) is our newly revamped rendering method that enhances interface performance. Now V7 can handle more annotations without slowing down. You can expect up to a 50x speed improvement, ensuring a smooth user experience and a refreshing rate of 50 FPS, even with thousands of annotations on a single file.

V7 Classes Tab Update

With this update, you now have a redesigned interface for managing classes within your teams and datasets. You can add various subtypes and control classes both at a global and local level. Assigning colors and custom hotkeys to your classes is now easier than ever.

More Features and Improvements

Feature
Feature
Feature
Feature
Performance
Performance
Performance
Performance
Fix
Fix
Fix
Fix
Fix
Fix
Fix


🚀 Added support for deleting classes that are used within annotate stage (Stage Specific Classes)

🚀 Performance improvements for polygon tool with LAYER_V2 (Accelerated Label Rendering)

🚀 Redesigned new classes page thumbnail editor

🚀 Logic Stage now supports multiple values

🚀 Added support for exporting Darwin Rasters into Semantic Masks

👾 Fixed annotation measurement label being shown regardless of the current frame index

👾 Fixed TAB behaviour about image manipulation preset selection triggering annotation selection

👾 Fixed wrong position for the class item contextual menu

👾 Fixed problem with skeleton updates overriding other skeleton classes

👾 Fixed OffscreenCanvas errors on older browsers

👾 Significantly reduced item loading time when the workview is opened if dataset items have large sizes

March 2023 Update

March 1, 2023
A series of cards from V7 Darwin

Logic Stages

These stages allow you to control your workflows with logical if/else conditions. The Logic stage checks for specific tags or annotations. Based on these, it can route your data to different stages. One logic stage can have multiple rules, which gives you a high degree of control. This feature is particularly useful for automating tasks related to annotation quality control. For example, specific tags (e.g. “blurry”) can route your image to an archive stage that removes low quality items from your workflow automatically, or to an additional review stage.

As of March, Logic Stage is behind a feature flag so if you’re interested in using it, reach out to us.

New Public Models

Since Text Scanner is the most popular pre-trained public model available in V7, we decided to fine tune it for specific tasks. There are three new document processing models available to all V7 users: Passport Scanner, Receipt Scanner, and Invoice Scanner. Now, you can quickly and accurately extract information from all sorts of documents. To access the models just head over to the Models tab of your V7 panel.

Consensus Stages

With consensus stages you can measure the overlap between two independent annotations. This feature is useful for determining the level of agreement between a human annotator and a model. You can also use it to compare two or more annotators and the accuracy of their labels. Alternatively, it is also possible to connect multiple models and use the consensus stage to measure their performance. You can specify the acceptable level of agreement and disagreement for each individual class.

Improved DICOM/NIfTI Experience

V7 now offers additional features to support medical image series and volumetric data. It is easier to adjust brightness and contrast with the Window Level Tool, import files into multi-series views, and change units to measure real size of lesions. You can use radiology-friendly controls to navigate volumetric data and NIfTI volume exports are now fully supported.

More Features and Improvements

Feature
Feature
Feature
Feature
Performance
Performance
Performance
Performance
Fix
Fix
Fix
Fix
Fix
Fix
Fix

👾 Fixed an issue that prevented the interpolation/keypoint options from appearing in the video timeline when selecting an annotation.
👾 Made a change to route non-worker roles to the Datasets page on sign-in, and workers are to the Workflows page
👾 Greatly improved brush performance for complex polygons with hundreds of segments.
👾 Fixed an issue preventing keyframes from appearing
👾 Added partner SSO credentials for all child teams when using Workspaces.
👾 Fixed an issue preventing polygons from merging when using the "Merge" option.
👾 Added additional error messaging when registering files from external storage using the API
👾 Fixed which kept polygons highlighted after being subtracted, making subtracting additional polygons clunky.
👾 Fixed an issue that prevented files from filtering by date modified
👾 Fixed the issue that was causing export generation to become stuck for some users.
👾 Fixed an issue with Model Stages which prevented them from running on .tif files.
👾 Fixed an issue where existing workflows containing a consensus stage could not be saved
👾 Fixed the issue causing dataset export links to not work as expected,
👾 Fixed the issue with workview timer logic. The timer will now pause after 60 seconds of inactivity, ensuring that all annotation time is accurately recorded.
👾 Fixed an issue where the usage metrics for all models were blank.
👾 Fixed annotation update issue that was preventing some annotation changes from saving.
👾 Added an improvement to ensure that users in child teams cannot log into their workspaces using a username and password if SSO configuration is shared from a parent team and SSO is enforced in the parent team.
👾 Fixed an issue that was preventing notifications from being sent out when users were tagged in comments using the comment tool.
👾 Fixed an issue where annotations remained on the canvas even when they were toggled off via the “Toggle Annotations” option.
👾 Fixed an issue where users belonging to multiple teams were experiencing an infinite loading page when trying to access an item from a different team.
👾 Fixed a bug where eraser tool didn’t work for polygons smaller than the eraser tip size.
👾 Fixed a bug that was causing performance issues with video playback when using the “,” and “.” shortcuts to scrub through video frames.
👾 Fixed a bug where selecting the table class, caused console errors and significant lag on the canvas.
👾 Fixed a bug where annotations with a persistence length greater than the length of the video were unable to be reduced in size.
👾 Renamed "Image file size" to "File size" in all dropdown menus.
👾 Fixed an issue where the delete export button was not shown in the Exports modal
👾 Improved page loading speed and reduced of intermittent display issues.
👾 Fixed bug causing clients to be unable to assign work to their workers.
👾 Developed new endpoint that enables unlinking a dataset from a workflow
👾 Added new feature which when enabled, will reduce quality of tiled images slightly in order to increase performance when using V7 with a slow connection.
👾 Fixed a bug where changing the size of frames on a video would render tags incorrectly.
👾 Fixed an issue where webhooks were not being sent after completing a stage.
👾 Fixed an issue where customers with workspaces on V7 2.0 were able to create child teams in 1.0.
👾 Fixed an issue where tabbing through annotations included hidden annotations instead of ignoring them as expected.
👾 Fixed an issue where files were not opening when workflow filters were applied.
👾 Uploading and Processing filters in the Dataset Management sidebar should be now visible for user by default.
👾 Fixed an issue where it was not possible to add attributes before a page refresh.
👾 Fixed an issue preventing some files from being opened when folder view was applied.
👾 Fixed an issue preventing images to appear in dataset instructions.
👾 Fixed an issue preventing Select All from working as expected in larger datasets
👾 Fixed an issue that displayed completed items in the workview for workers
👾 The brush tool now periodically persists data to the server during a create or edit. This prevents unsaved annotations if an annotation is made for a very long time without saving and the annotator leaves the browser.
👾 Eliminated a memory leak when navigating back and forth from the workview to the data page.

November 2022 Update

November 22, 2022
A series of cards from V7 Darwin

Introducing Webhook Stages

We’re continuing to make it easier to structure your annotation flows and to collaborate more effectively between teams. We’ve added an additional workflow stage to the workflow creation wizard as part of the improvements we’re adding to the 2.0 version of V7. The Webhook Workflow Stage now enables users to send webhooks at any stage of the workflow regardless of its state. These webhooks also include useful annotator information. For example, if you want to send Slack notifications when annotations reach a certain stage, this is now possible through Webhook Stages. We’ve also simplified the workflows API to enable simpler programmatic creation, editing, and deletion of workflows and their stages.

Bring Your Own Model

Our recently released Bring Your Own Model (BYOM) feature now allows you to bring the models that you host on your infrastructure and use them within the V7 platform. Three ways you can become a power user via BYOM:

  • Add a Model Stage to your Workflows for automatic inference.
  • Use your model in the Workview with the Auto-Annotate tool.
  • These models behave identically to models trained using V7’s AutoML to enable users to feed data to models as you would models trained on V7.

Further support for NIfTI format

We're continuing our work to make V7 a leading solution for medical annotation teams. We're pleased to announce the rollout for the support of NIfTI file formats and support for volumetric data within V7’s annotation platform. This includes:

  • Support for annotation on NIfTI file formats without losing metadata.
  • Extract Axial, Coronal, and Sagittal planes to view within a 3D space through the 'Extract Views' option when importing NIfTI and DICOM files.
  • Ability to export annotations in NIfTI (coming soon).
  • Reference lines are rendered at the annotation layer.

More Features and Improvements

Feature
Feature
Feature
Feature
Performance
Performance
Performance
Performance
Fix
Fix
Fix
Fix
Fix
Fix
Fix

👾 Fixed bug where the archive/restore button on a single item in a v2 dataset was not functioning

👾 Resolved the issue of being unable to open files after turning off folder view, allowing for smooth file accessibility.

👾 Addressed the problem of useRoute/useRouter breaking, which was causing navigation links in the app to malfunction

👾 Set the child team limit to be the maximum number of credits available in the Parent team by default, ensuring efficient resource allocation

👾 Added more permissions to workforce managers, expanding their capabilitiesAdded more permissions to workforce managers, expanding their capabilities

👾 Strengthened instructions processing against XSS

👾 Ensured desynced video audio streams are processed correctly

👾 Increased timeout for downloading files from Azure and reducing download errors

👾 Ensured trained model and training sessions scroll correctly

👾 Implemented a timer reset after 60 minutes

👾 Fixed the issue of not updating UI when sub-annotation is changed

👾 Utilized the dataset slug as part of export storage key

👾 Removed the instance_id ‘hotfix’, streamlining the codebase and improving overall stability

👾 Enabled Darwin JSON 2.0 for all new V2 teams, providing a more advanced and efficient data management experience

👾 Fixed JWT signatures for dataset instructions, ensuring secure and reliable data handling

June 2022 Update

June 20, 2022
A series of cards from V7 Darwin

V7 2.0

V7 just got an upgrade. We’ve refactored our workflow system and API and are excited to announce what we’ve internally dubbed V7 2.0. We will be rolling out 2.0 to eligible teams teams over the coming development cycles to ensure a smooth transition from one version to the next. Here’s what’s in store:

Workflows 2.0

Workflows can be built from scratch, or duplicated from templates in the new workflow creation wizard. This release is just the beginning - we’ve built out the foundation that will allow us to add entirely new stage types like Logic stages and Webhook stages in our upcoming releases, giving you and your team the ability to add custom logic and create more dynamic workflows.

An All-New API

We’ve rebuilt V7’s API, adding new endpoints and new functionality while maintaining backwards-compatibility with our SDK. These changes pave the way for programatically managing projects to be easier than ever, and will give us the ability to support an even broader array of filetypes and annotation tasks in future releases.

Unprecedented Performance

Dataset management, bulk tasks, and load times will all be getting noticeable boost on V7 2.0. New logic for handling items adds speed to every part of the system, and lays the foundation for more impactful performance improvements with each upcoming development cycle.

Brand New Documentation

Our guides and API documentation are now versioned. Switch between v1.0 and v2.0 to see what's changed. Teams will get advance notice before they are migrated to 2.0, and all endpoints available upon launch can be reviewed here. 

More Features and Improvements

Feature
Feature
Feature
Feature
Performance
Performance
Performance
Performance
Fix
Fix
Fix
Fix
Fix
Fix
Fix
Fixes

👾 Fixed a bug that prevented attributes from rendering without a refresh.

👾  Pushed a security patch to prevent unauthenticated users from viewing external storage names.

👾 Adjusted frame rate handling to prevent some videos from importing to an Error status.

👾 Updated assignment rules to allow parent team workers to self-assign files in child teams.

👾 Fixed an issue that prevented users from being able to delete exports.

👾 Fixed an issue that caused the previously-selected annotation from pasting when when copying & pasting.

👾 Fixed a display issue causing annotations to shake when moving them with the arrow keys.

👾 Fixed rendering of annotations in hanging protocol DICOMs.

👾 Fixed an issue that caused polylines to pan incorrectly on the image canvas.

👾 Fixed an issue that prevented files containing graph annotations from panning.

👾 Moved the position of keypoint labels to prevent them from obscuring keypoints.

👾 Fixed an issue causing labels to be incorrectly capitalised.

👾 Fixed an issue preventing table classes from being deletable.

👾 Fixed an issue preventing changes in high-density files from being saved when annotating with a slow internet connect.

👾 Fixed an issue preventing Pascal VOC annotations from exporting.

👾 Fixed an issue that prevented all available filetypes like PDFs and videos from being useable in the Instructions section of the dataset Settings page.

👾  Fixed an issue with API key permissions that prevented attributes from being added.

👾 Made generic Auto Annotate the default model equipped when selecting the model tool.

👾 Fixed an issue that prevented attributes from being added from the Annotations menu in the workview.

👾 Fixed an issue that caused tags created from within the workview to have attributes enabled automatically.

👾 Fixed an issue that caused tags to be duplicated across the different files within a hanging protocol layout.

👾 Changed how interpolation is handled by complex polygons (polygons with holes or islands) to smooth out polygon interpolation across frames.

👾 Fixed the logic behind how polygon vertices are added to complex polygons to prevent them from being added to the wrong section.

👾 Fixed an issue that prevented credits from being pulled down from parent teams into child teams.

👾 Fixed a display issue that caused an infinitely-spinning wheel when correcting an Auto-Annotate generated polygon.

👾 Fixed a similar issue that caused a green background to appear behind the spinner when using Auto-Annotate.

👾 Generating an export by the current filters in a dataset now respects negative filters.

👾 Fixed some small discrepancies between the datasets page and workflow filters.

👾 Fixed an issue where polygons were temporarily disappearing or not selectable when editing with the brush tool.

👾 Fixed an issue that prevented tags from being added to the first image in a hanging protocol layout.

👾 Improved slow buffering/load times for video and DICOM files.

👾 Fixed a report generation timeout bug that prevented annotator reporting data from being visible.

👾 Fixed an issue that caused the brush tool to invert the highlighted sections of complex polygons.

👾 Added hours to the timeout limit for processing files to prevent files from appearing stuck in Processing indefinitely

February 2022 Update

February 28, 2022
A series of cards from V7 Darwin

Introducing: The New Model Tool

Auto-Annotate is getting a major power-up: Use any model to automatically label polygons, bounding boxes, text, and tags. The Model Tool lets you map classes to models from the internet, and will surface confidence scores for every label created.

Confidence scores within the Annotation interface

See confidence scores for every annotation, spot ambiguous objects and set confidence scoring thresholds to control the accuracy of your automatically-labelled data. 

Support for all model types 

Whether it's one of your team's trained object detection or classification models, or one of our public models like the text scanner, the Model Tool can output any annotation type. You may even use it at lower complexity levels, for example use an instance segmentation model to produce bounding boxes.

Label specific regions or entire files

Create multiple unique labels at once by drawing a bounding box over several objects, or run a model over an entire file by hitting Cmd + Enter.

Model Tool also works on ultra-large files, medical imaging, and PDFs.

Triage hard tasks with AI Models 

One of our favorite ways of using Model Tool is to pre-process data and flag low-confidence boxes. This is a great way to flag out-of-distribution objects and direct reviewers to them. You may also use multiple model versions to compare their score on the same image.

More Features and Improvements

Feature
Feature
Feature
Feature
Performance
Performance
Performance
Performance
Fix
Fix
Fix
Fix
Fix
Fix
Fix

New Features

🆕 Label segments can now be easily moved/resized in video timelines when annotating on touch displays.

Performance Improvements

🚀 We've made zooming and panning in multiple-gigabyte TIFF files significantly smoother.

🚀 DICOM series and their slices now have significantly-improved loading times and playback speeds.

Fixes

👾 New annotations now always appear at the top of the Annotations menu.

👾 Corrected the alignment of rendered frames in streamed videos with the frame in the video timeline.

👾 Fixed an issue that prevented Auto-Annotate from functioning in videos with streaming enabled.

👾 Corrected the interpolation algorithm for Ellipse annotations to ensure smooth interpolation between key frames.

👾 Fixed an issue that caused the layering of annotations in the Annotations menu to be inverted when compared to the layers in the exported JSON file.

👾 Tweaked interpolation behaviour to prevent complex polygons (polygons with multiple parts) from disappearing in interpolated frames.

👾 Files in a Complete state no longer display any annotation tools to reflect the fact that they are uneditable.

👾 Fixed an error that prevented DICOMs with hanging protocol layouts from being edited.

👾 Class filters now reset when moving between datasets.

👾 All class names were previously capitalised in the workview. This is no longer the case.

👾 Keypoints now change appearance when hovering over them with the cursor. This makes them clearer when not the cursor is not hovering above them.