Fixless Audits

Fixless Audits is a powerful feature designed to enhance the quality and accuracy of task auditing. Unlike regular audits, fixless audits allow auditors to submit their feedback, without the need to perform any fixes, making the auditing process much quicker and effective.

This type of audits can be created both automatically, via API, and manually through the LidarLite tool.

This document aims to provide a step by step guide on how to review the API generated audits, and also how to create an audit manually using the tool.

Create a Fixless Audit

Audits created with it are composed of Feedback Item (FI)s, which indicate errors, comments, confirmations, or flags on tasks.

Within the Scale platform, Feedback Items can be dynamically updated. Feedback Item states (accepted, rejected, disputed) can be modified as contributors address the feedback. This flexibility ensures that the auditing process is both thorough and responsive.

Error-type Feedback Items, which are used to calculate quality scores. Only the latest Fixless Audit for each task is considered. However, all historical audits for a task are accessible via the API and can be reviewed in the Scale Auditing tool.

By indicating the “Scope” of the FI auditors can defines the relevant component of the task (e.g., annotation_id, attribute, frame interval). Scope also includes an error category field.

Multiple Fixless Audits can be submitted for the same task, but only the most recent audit affects the quality score. This system ensures that the most up-to-date and accurate information is used for evaluation.

Additionally, Fixless Audits submitted via the API can be modified and adjusted in the Scale Auditing tool, allowing auditors to add, delete, or edit Feedback Items as needed.

Below is a detailed breakdown of the different types, states, and categories of Feedback Items, along with their respective fields and metadata:

Feedback Item	Type	Is Required	Options
Type: Type is determined at time of primitive creation, but can be changed in subsequent audits The type used will determine whether certain fields are required	String	Yes	Error : The annotation is incorrect Flag : The annotation needs to be reviewed by an auditor Confirmation : The annotation is correct Comment : Generic feedback, does not affect quality scores
State : States can only be changed by Scale users, and can be changed multiple times	String	Yes	Open : Default state when the primitive is created Accepted : Selected when Scale agrees with the feedback item Disputed : Selected when Scale disagrees with the feedback item Rejected : Selected if the customer agrees with Scale on a disputed feedback item and removes the Feedback Item from the quality score calculation
Scope	Object	Yes	For more detail on the different type of scopes and their params, check the table below.
Category Fields: Error Category	String	Yes, only for type “error“	Geometry error Position error Orientation error Label error Attribute error Extraneous error Missing error
Category Fields: Error Severity	String	Yes, only for type “error“	Mild : Used when the auditor wants to flag an error, but the error doesn’t exceed our SLA error threshold (e.g. cuboid position is off by <30cm) These errors will not affect the quality score Standard: Used when the auditor wants to flag an error that exceeds our SLA error threshold (e.g. cuboid position is off by >30cm) These errors will affect the quality score Severe : Used when the auditor wants to flag an error that exceeds our SLA error threshold (e.g. cuboid position is off by >30cm) AND the error is a critical error that requires special attention These errors will affect the quality score
Description	String	No	A free-form text area where users can enter any relevant information or details.
Metadata	Object	No

Request

POST

// --- /v1/audits --- //

// PUT /v1/audits
body FixlessAuditCreate
201 response FixlessAudit

// GET /v1/audits?id=xxx&task_id=yyy
200 response FixlessAudit[]

// --- /v1/audit --- //

// GET /v1/audit/:id
200 response FixlessAudit

Creation Payload

FixlessAuditCreate {
  type: 'fixless';
  result: 'accepted' | 'rejected';
  task_id: string;
  comments?: string;
  // from task.response.annotations.url
  target_response_url: string;
  feedback_items?: FeedbackItemCreate[];
  metadata?: { [key: string]: any };
}

FeedbackItemCreate {
  type: 'comment' | 'error' | 'flag' | 'confirmation';
  scope: FeedbackItemScope; // see DTO
  category?:
    | 'attribute'
    | 'extraneous'
    | 'geometry'
    | 'label'
    | 'missing'
    | 'position';
  severity?: 'mild' | 'standard' | 'severe';
  description?: string;
  metadata?: { [key: string]: any };
}

Feedback Item Scope

FeedbackItemScope =
  | FeedbackItemScopeAnnotation
  | FeedbackItemScopePolygon
  | FeedbackItemScopePoint
  | FeedbackItemScopeScene;

// Most errors should use this scope
FeedbackItemScopeAnnotation = {
  type: 'annotation';
  annotation_id: string;
  

// Defaults to the scene's primary sensor (typically the first lidar)  // Typically used to identify a specific camera for 2D projections
sensor_id?: string | number;

  // attribute name. defined iff scoped on an annotation attribute

  attribute?: string;
  interval: FeedbackItemTimestampRange;
}

// Only for missing annotation errors
FeedbackItemScopePolygon {
  type: 'polygon';
  vertices: [[x0, y0], [x1, y1], [x2, y2], ...];
  // defined iff scoped on a 2d projection
  camera_id?: string;
  interval: FeedbackItemTimestampRange;
}

// Only for missing annotation errors
FeedbackItemScopePoint {
  type: 'point';
  coordinates: [x, y];
  // defined iff scoped on a 2d projection
  camera_id?: string;
  interval: FeedbackItemTimestampRange;
}

FeedbackItemScopeScene {
  type: 'scene';
  // attribute name. defined iff scoped on a scene attribute
  attribute?: string;
  interval: FeedbackItemTimestampRange;
}


// INTERVAL

FeedbackItemInterval = {
  type: 'frame_range';

  // defaults to time since start of scene
  epoch?: 'scene' | 'unix';

  // frame index
  start: number;
  end: number;

};

Data Transfer Objects (DTO)

FixlessAudit {
  id: string;
  // SRNs can be used in place of id on endpoints
  srn: 'srn:scale:avcv:audit:{{id}}'
  type: 'fixless';
  result: 'accepted' | 'rejected';
  task_id: string;
  comments?: string;
  target_response: { url: string };
  feedback_items?: FeedbackItem[];
  metadata?: { [key: string]: any };
  active: boolean; // whether this is the latest audit
  source: 'api' | 'lidarlite' | 'classic';
  created_by: string;
  created_at: iso_date_string;
  updated_at: iso_date_string;
}

FeedbackItem {
  id: string;
  type: 'comment' | 'error' | 'flag' | 'confirmation';
  state: 'open' | 'accepted' | 'disputed' | 'rejected';
  scope: FeedbackItemScope;
  category?:
    | 'attribute'
    | 'extraneous'
    | 'geometry'
    | 'label'
    | 'missing'
    | 'position';
  severity?: 'mild' | 'standard' | 'severe';
  description?: string;
  metadata?: { [key: string]: any };

  // updates to the state field (rest are immutable)
  updated_at: iso_date_string;
  updated_by: string;
  grader_output: {
    error_weight?: number;
    explanation?: string;
    conflict: boolean;
  }
}

FixlessAuditConflicts {
  auditConflictStatus: 'accepted' | 'accepted-with-conflicts';
  rejectedFeedbackItems: FeedbackItem[]; 
}

Example

import requests
from requests.auth import HTTPBasicAuth

headers = {"Accept": "application/json"}
auth = HTTPBasicAuth('{{ApiKey}}', '') # No password

response = requests.request(
    "PUT",
    "https://api.scale.com/v1/audits",
    json={
        "type": "fixless",
        "result": "rejected",
        "task_id": "task_id_here",
        "comments": "test reject comment",
        "target_response_url": "value from task.response.annotations.url",
        "feedback_items": [
            {
                "type": "comment",
                "scope": {
                    "type": "annotation",
                    "annotation_id": "abc123",
                    "sensor_id": "789xyz",
                    "interval": {
                        "type": "frame_range",
                        "start": 0,
                        "end": 10,
                    },
                },
                "severity": "mild",
                "description": "feedback 1",
            },
            {
                "type": "confirmation",
                "scope": {
                    "type": "scene",
                    "attribute": "weather",
                    "interval": {
                        "type": "frame_range",
                        "start": 0,
                        "end": 100,
                    },
                },
                "description": "feedback 2",
            },
        ],
        "metadata": { "test": True },
    },
    headers=headers,
    auth=auth,
)

response.status_code # 201

response.text
{
  "id": "6594cfca95644b72685e76e1",
  "type": "fixless",
  "task_id": "task_id_here",
  "task_type": "sensorfusion",
  "result": "rejected",
  "target_response": {
    "url": "value from task.response.annotations.url"
  },
  "comments": "test reject comment",
  "feedback_items": [
    {
      "id": "6594cfca95644b72685e76e5",
      "type": "comment",
      "state": "open",
      "scope": {
        "type": "annotation",
        "annotation_id": "abc123",
        "sensor_id": "789xyz",
        "interval": { "type": "frame_range", "start": 0, "end": 10 }
      },
      "severity": "mild",
      "description": "feedback 1",
      "updated_at": "2024-01-03T03:08:58.200Z",
      "updated_by": "user_id_123456"
    },
    {
      "id": "6594cfca95644b72685e76e6",
      "type": "confirmation",
      "state": "open",
      "scope": {
        "type": "scene",
        "attribute": "weather",
        "interval": { "type": "frame_range", "start": 0, "end": 100 }
      },
      "description": "feedback 2",
      "updated_at": "2024-01-03T03:08:58.201Z",
      "updated_by": "user_id_123456"
    }
  ],
  "metadata": { "test": true },
  "active": false,
  "source": "api",
  "created_by": "user_id_123456",
  "created_at": "2024-01-03T03:08:58.162Z",
  "updated_at": "2024-01-03T03:09:42.943Z"
}