Per-Sample Dice

Available

semantic_segmentation.per_sample_dice

Dice overlap for each image after semantic masks are decoded to dense arrays.

Contract

Field	Value
version	1.0.0
metric_type	pointwise
scorer_contract	per_row
direction	higher_is_better
entrypoint	aip_metrics_semantic_segmentation.answered.per_sample_dice
target_kind	None

Required columns

• image_id
• label
• predictions

Accepted schemas

[
  {
    "name": "gdi_image_v1",
    "task_types": [
      "semantic_segmentation"
    ]
  }
]

Methodology

1. Decode the row's ground truth to an integer class map and prediction to one-hot channels.
2. Compute cv-eval micro Dice across all pixels and channels in that image.

Score semantics

Scores range from 0 to 1. Higher is better; 1 means perfect pixel overlap for the image.

Worked example

If twice the intersecting pixel count is 12 and the total predicted-plus-true area is 16, Dice is 0.75.

Configuration schema

{
  "properties": {
    "inverted": {
      "default": false,
      "title": "Inverted",
      "type": "boolean",
      "x-aip-param-role": "scoring_metadata"
    },
    "max": {
      "default": 1.0,
      "title": "Max",
      "type": "number",
      "x-aip-param-role": "scoring_metadata"
    },
    "metric_name": {
      "default": "per_sample_dice",
      "title": "Metric Name",
      "type": "string",
      "x-aip-param-role": "scoring_metadata"
    },
    "min": {
      "default": 0.0,
      "title": "Min",
      "type": "number",
      "x-aip-param-role": "scoring_metadata"
    },
    "smooth": {
      "default": 0.001,
      "minimum": 0.0,
      "title": "Smooth",
      "type": "number"
    },
    "threshold": {
      "default": 0.5,
      "maximum": 1.0,
      "minimum": 0.0,
      "title": "Threshold",
      "type": "number"
    },
    "weight": {
      "default": 1.0,
      "title": "Weight",
      "type": "number",
      "x-aip-param-role": "scoring_metadata"
    }
  },
  "title": "InitializationConfig",
  "type": "object"
}

Execution

{
  "emits_metric_family": false,
  "function_name": null,
  "max_concurrency": null,
  "processing_kind": "cpu",
  "stream_batch_size": null,
  "timeout_seconds": 600
}

Complete manifest

accepts:
- name: gdi_image_v1
  task_types:
  - semantic_segmentation
config_schema:
  properties:
    inverted:
      default: false
      title: Inverted
      type: boolean
      x-aip-param-role: scoring_metadata
    max:
      default: 1.0
      title: Max
      type: number
      x-aip-param-role: scoring_metadata
    metric_name:
      default: per_sample_dice
      title: Metric Name
      type: string
      x-aip-param-role: scoring_metadata
    min:
      default: 0.0
      title: Min
      type: number
      x-aip-param-role: scoring_metadata
    smooth:
      default: 0.001
      minimum: 0.0
      title: Smooth
      type: number
    threshold:
      default: 0.5
      maximum: 1.0
      minimum: 0.0
      title: Threshold
      type: number
    weight:
      default: 1.0
      title: Weight
      type: number
      x-aip-param-role: scoring_metadata
  title: InitializationConfig
  type: object
dependencies: []
description: Per-image Dice overlap between predicted and ground-truth semantic masks.
direction: higher_is_better
display_name: Per-Sample Dice
entrypoint: aip_metrics_semantic_segmentation.answered.per_sample_dice
execution:
  emits_metric_family: false
  function_name: null
  max_concurrency: null
  processing_kind: cpu
  stream_batch_size: null
  timeout_seconds: 600
kind: metric
manifest_version: '1'
max_prompt_slots: 0
metric_metadata:
  methodology:
  - Decode the row's ground truth to an integer class map and prediction to one-hot
    channels.
  - Compute cv-eval micro Dice across all pixels and channels in that image.
  score_semantics: Scores range from 0 to 1. Higher is better; 1 means perfect pixel
    overlap for the image.
  summary: Dice overlap for each image after semantic masks are decoded to dense arrays.
  worked_example: If twice the intersecting pixel count is 12 and the total predicted-plus-true
    area is 16, Dice is 0.75.
metric_type: pointwise
name: semantic_segmentation.per_sample_dice
partition_types: []
required_columns:
- image_id
- label
- predictions
required_kinds: []
scorer_contract: per_row
target_kind: null
unsupported_trace_shapes: []
version: 1.0.0