Uploader UI
- Choose Annotation Type
- Classification (whole image labels)
- Bounding Boxes (rectangles)
- Polygon Segmentation (precise outlines)
- Point Annotation (specific points)
- Define Labels
- Add/remove label names
- Optional descriptions for each label
- Clear explanations of what each label means
- Third: Set Instructions
- Detailed instructions for annotators
- Guidelines for edge cases
- Example annotations
- Quality requirements
- Helper text for writing good instructions
- Upload Images
- Bulk upload interface
- Preview thumbnails
- File validation
- Progress tracking
Annotator UI
- Adaptive Tools:
- Left toolbar changes based on annotation type (boxes/polygons/points)
- Undo/redo functionality
- Zoom controls
- Clear visual feedback for active tools
- Work Area:
- Large central canvas
- Image zoom and pan controls
- Annotation overlay rendering
- Previous/Next image navigation
- High contrast background for better visibility
- Right Sidebar:
- Label selection with color coding
- Project instructions always visible
- Keyboard shortcuts reference
- List of annotations on current image
- Easy label switching
- Top Bar:
- Time tracking
- Earnings information
- Progress tracking
- Issue reporting
- Task submission
- Quality Features:
- Clear task instructions
- Keyboard shortcuts for efficiency
- Issue flagging system
- Progress tracking
- Easy access to all needed tools
User flow for each data type
Common features across all types
Quality Control:
- Each image should be annotated by 2 different workers
- Agreement checks between annotators
- Random quality check images
- Time per task monitoring
Keyboard Shortcuts:
- Next/Previous Image: Arrow keys
- Delete Last Annotation: Delete/Backspace
- Quick Label Selection: Number keys 1-9
- Zoom: Ctrl + Mouse wheel
Support Tools:
- Brightness/Contrast adjustment
- Zoom in/out
- Issue flagging
- Reference image examples
- Annotation guidelines
Classification
Customer Setup:
- Define list of possible labels (e.g., “Dog”, “Cat”, “Bird”)
- Specify if single-label (choose one) or multi-label (select all that apply)
- Upload batch of images
Annotator Workflow:
- Image appears in workspace
- Review list of available labels
- Select appropriate label(s) from right sidebar
- Optional: Add notes if uncertain
- Submit and move to next image
Bounding Boxes
Customer Setup:
- Define objects to be detected (e.g., “Car”, “Person”, “Sign”)
- Specify if overlapping boxes are allowed
- Upload images
- Optional: Define minimum object size
Annotator Workflow:
- Select label from right sidebar (e.g., “Car”)
- Click and drag to draw rectangle around object
- Adjust box corners if needed
- Repeat for all instances in image
- Can delete/redraw boxes if needed
- Submit when all objects are marked
Polygon Segmentation
Customer Setup:
- Define objects needing precise outlines (e.g., “Building”, “Road”, “Tree”)
- Specify required precision level
- Upload images
- Optional: Set minimum vertex count
Annotator Workflow:
- Select label from sidebar
- Click points to create polygon vertices
- Connect back to first point to complete shape
- Can adjust vertex positions
- Right-click to finish current polygon
- Repeat for all instances
- Submit when all objects are segmented
Point Annotation
Customer Setup:
- Define points of interest (e.g., “Facial Landmarks”, “Joint Positions”)
- Specify if points need to be in specific order
- Upload images
Annotator Workflow:
- Select point type from sidebar
- Click precise location on image
- Can adjust point position by dragging
- Follow ordered sequence if required
- Delete misplaced points with right-click
- Submit when all required points are placed
Customer Workflows
For specialized needs, can combine multiple types:
- Bounding box + Classification (e.g., mark cars and their colors)
- Points + Lines (e.g., mark joints and connect them)
- Polygon + Attributes (e.g., outline buildings and mark their type)
These require additional setup but use the same basic tools.