Avala’s query language lets you filter dataset items, annotations, and export results using structured expressions. Use it in the Mission Control search bar, export filters, and the filter_query_string parameter in the Exports API.
Operators
| Operator | Description | Example |
|---|
= | Equal to | annotation.label = "car" |
!= | Not equal to | annotation.label != "unknown" |
> | Greater than | annotation.attribute.confidence > 0.9 |
< | Less than | annotation.attribute.confidence < 0.5 |
>= | Greater than or equal to | annotation.attribute.area >= 100 |
<= | Less than or equal to | annotation.attribute.area <= 500 |
Logical Operators
Combine multiple conditions with AND, OR, and NOT. Use parentheses to control grouping.
| Operator | Description | Example |
|---|
AND | Both conditions must be true | annotation.label = "car" AND annotation.attribute.occluded = "false" |
OR | Either condition must be true | annotation.label = "car" OR annotation.label = "truck" |
NOT | Negates a condition | NOT annotation.label = "unknown" |
( ) | Groups conditions | (annotation.label = "car" OR annotation.label = "truck") AND annotation.attribute.truncated = "false" |
Annotation Queries
Filter by annotation properties.
By Label
annotation.label = "car"
annotation.label != "pedestrian"
By Attribute
Query annotation attributes using dot notation:
annotation.attribute.occluded = "true"
annotation.attribute.truncated = "false"
annotation.attribute.confidence > 0.8
By Annotation Type
annotation.type = "bounding_box"
annotation.type = "polygon"
annotation.type = "cuboid"
metadata. prefix followed by the field name:
metadata.weather = "rainy"
metadata.scene_type = "highway"
metadata.time_of_day = "night"
Metadata field names are case-sensitive and must match the exact field name used when the metadata was uploaded.
Slice Queries
Filter items that belong to a specific data slice:
slice = "training"
slice = "validation"
slice = "edge-cases"
Reference ID Queries
Filter items by their reference ID:
ref_id = "frame_00123"
ref_id = "scene_042_cam_front"
String Quoting
Use double quotes around values that contain spaces or special characters:
annotation.label = "traffic light"
metadata.location = "San Francisco"
slice = "hard examples"
annotation.label = car
annotation.label = "car"
Examples
Find all items with car annotations:
Find items with high-confidence annotations:
annotation.attribute.confidence >= 0.95
metadata.weather = "rainy" AND metadata.scene_type = "highway" AND slice = "training"
(annotation.label = "car" OR annotation.label = "truck") AND annotation.attribute.occluded = "false"
NOT annotation.label = "pedestrian"
ref_id = "frame_00500" AND annotation.type = "cuboid"
