Release 6.10
Changelog for Clarifai Release 6.10
Scribe
When a reviewer opens the "For Review" tab, determine whether there is work to be reviewed
Remove Labeler V1
Keybindings not working when buttons are clicked
Interpolation transforms not working
Update Search API Endpoint
Prediction bounding box shows for concepts not included in task
Image filters don’t work
AI Assist should not create predictions with 0% confidence
AI Assist is creating annotation but is not auto marking the green check box or the red "X"
AI Assist sends incorrect app ID when fetching workflows
Full assignment strategy results in error during review
Tests: Labeler Interpolation Sagas
Tests: Labeler Inputs Sagas
Tests: Labeler Canvas Sagas
Show Sentry Event Id for Saga Errors
Labeler AI Assist Breaking for custom User ID
Unable to load/access "Assigned to me" labeler tasks
Only able to draw one bounding box while using interpolation
Brightness, Saturation and Invert doesn't change
Task Edit form doesn't show the correct FPS value from the task object
Task Form shows warning when creating a new task
Place Order button remains on page when switching back from LaaS option to self-labeling
Tests: Labeler UI Sagas
When "My own Labelers" is clicked. All fields are not shown back
Testing LabelOrder reducer - user journey
Change cancel order copy
Unit test LabelOrder reducer.
Testing LabelOrder reducer
Feature gate Labelling suggestions
Pinning apps breaking search app functionality
Unable to create a new Task in Labeler
Better Video Features
Opening a task should ONLY fetch annotations for the task that was opened
Task Edit form doesn't pre-populate worker strategy selection
Submit for Review logic incorrectly assumes the last input is last
On the Task creation form Add section to Review strategy block for Consensus type
Fix key issue in ExplorerAnnotationStatusList component in Explorer
Handle Noisy Labels
Optimistically update "don't show again" in onboarding modal of Task Form
Flipkart would like better formatting of markdown for instructions
Show onboarding modals for each persona: Task creator, Reviewer, Worker
Onboarding tooltips for /apps doesn't go away when I click "awesome"
Pressing ESC patches my metadata 4 times in a row
Move Hubspot chat icon into the navbar
Task Form Inputs and Buttons are Misaligned
Task Creation submit not redirecting
In Labeler, “Skip” and "Reject" buttons are hidden by the questions bot blob
Toggling panning makes the image blurry
Clicking "Create Task" doesn't force you to exit the task creation screen.
App Details - Tool Tip only shows 1/3 (there doesn't seem to be a 2/3 or 3/3)
Display onboarding modal for Reviewer
Display onboarding modal for Task Creator
Review Strategy for TaskForm
AI Assist for TaskForm
Manual reviewers for TaskForm
Worker strategy input for TaskForm
Worker input for TaskForm
Create ProposerWorkflow input for TaskForm
Create VideoFps input for TaskForm
Create Concepts input for TaskForm
Create InputSource input for TaskForm
Create TaskType input for TaskForm
Create separate styled component file for FormTypeSection
Onboarding modal points to 2 videos and they link deeper into the same video.
Enlight
Implement track association module
implement log probability using kalman state as input
cannot save model details in portal with lrate < 0.001
[OCR] Scene Text Reco & Text Classify Integration
Splice Concepts Together
New Layers / Arch
Models on New Data Types
Neural Net Toolkit
Recurrent Nets Up and Running
YOLO dataset script
deploy landmark model
Update roll up for task_input_counts table to have worker_id information
Implement annotation count
Integrate filter, status, association into a single tracker model
Implement aggregator
Design simulated data that mimic low confidence / high recall detections
Design training losses
Try box in x y a r format (SORT format)
Try box in x y w h format
Checkpoint for test eval can be deleted before eval is started
Improve face orientation bucketing of LandmarkAlignTransform in DataProvider
Use tracker state in TrackerPredict calls
Make Text Aggregator Operator public
Remove transformers dependency
Simple visual tracker example
Ingest VGGFACE2
Explore multi-language support
Rewrite Export Scripts
Introduce region threshold model
Add filter_other_concepts param to concept threshold model
Remove the unnecessary requires_sequential_frames field of ModelType
Add support for image crop model to work on video
Fix duplicate_app.go restriction on number of annotations
Notebook to demo tracker evals
RNN vs Model-based tracking: Real data
Armada
Triton GRPC Client Upgrade
Migrate Spire to use TextTokenizer Transform
Implement TextTokenizer Transform
Export model to TorchScript
HDFS for EID storage
Experiment tracking improvements
Trace from EID
EID management
Ingest A Video dataset
Resolve TRT saved model Issues Across GPU compute Capabilities
People/Vehicle Detector TRT conversion and inference benchmarking
Portal
CSS on toggles within model mode was cutting off the toggle
Model creation broken
Create Hubspot dashboard with funnels and charts
Capture customer behavior analytics
CORS error prevents uploaded images from rendering
Properly configure S3 bucket for inputs to allow portal (prod and dev)
Investigate why the vendors bundle is 12MB
Render relation-based parent-child relationship in Proposals
Proposers UI with a new annotations reducer
Create annotations module in store/explorer for Proposers UI
Move Proposers data transformation to a saga+web worker
Integrate Proposers with Classification apps
Show App Concepts alongside model concepts in Proposers UI
Explorer does exponentially increasing duplicate calls to GET inputs endpoint
Introduce knowledge graph link creation when check/X is clicked on clarifai/main models
update created-at timestamp format to: hours/minutes/seconds for model-input
data-mode - uploaded-images multiple elements get re-rendered
Zoom-in & zoom-out controls for explorer geo
Allow users to upload the same url twice from data mode
Set app-id instead of the app-name on new app-creation for readable urls
Fully qualified url for app-listing
NaN being displayed in model-versions
Portal does not parse links properly
Unable to view versions or eval metrics page as a collaborator
Data Ingestions
Create initial eval script to test cascade approach
Work with BE to nurse input uploads
Get Input Embeddings and Cluster over All
Get Input Embeddings and Cluster over Concepts
Data Mode page performance issues when pasting 60 inputs in production
Bulk unlabeling doesn't work
Change model type to model type ID everywhere
Slider for concept threshold should allow for full precision floats with box to set the value
Upload third party models through Portal
Develop the model predict view for portal for existing models.
Add dropdown for model_type_id
Allow users to upload pre-trained models
Make versions tab as default in model details page
Update ModelPredictions component to make use of reselect & sagas
Create collectors always selects user as clarifai
AnimatedList caches the children nodes causing improper renders
Sending invalid fields during model creation
Workflow page crashes
Improve performance of explorer
Remove console errors in various parts of portal
GetInputCount() is called on loop after exiting data mode
Create new Collector redirects to apps page
Collectors Adding app model bug
copy public workflow do not populate workflow model
In eval metrics page, "Concept by Concept" section shows counts that are incorrect
In eval metrics page, "1 split" section is incorrect
Prevent patching model annotations
Deprecate old search endpoint from FE
Add Embed Model Version Id to Bulk Labeler Classification Annotations
Range Slider on Eval Metrics Page causes app crash when moving
Allow copying public workflows into existing workflows
API
Overwrite user annotation info with annotation writer model info
Unable to add polygon annotation with embed_model_version_id
Experiment with docs.clarifai.com using API versions
Make gRPC Swift client
Can't assign a task to myself if I have a custom user_id
Annotation writer model & concept thresholder create incorrect additional annotations
App copy doesn't stop after timeout
Asset search by status
Monitor media processor predict err rate
Log spire name
Clean idle/pending asset publish video asset to pipeline separately
Last updated
