Back to Swatches

CrowdAI - Automated Identification of Dented Cans

Fabric Score 3.6
PrivateManufacturingCrowdAINon-GenAI modelExecutive / Org Leadership ×1

Workflow Diagram

CrowdAI - Automated Identification of Dented Cans workflow diagram

Fabric Score

ValueEfficacySecurityRiskExternalitiesEfficiency3.84.03.73.33.43.83.6Fabric Score

Task

On the manufacturing line, there are video cameras that are placed on intermittent areas of the floor. Using video camera footage to identify dented aluminum cans.

Intent

The AI system automatically identifies dented cans on the manufacturing line.

AI Workflow

Input

On the manufacturing line, video data is collected.

Process

AI system automatically creates boxes and locations of dented cans using a computer vision system.

Output

Manufacturing factory floor analysts see the overlaid boxes with timestamps of where the dented cans exist. Output is timestamp, certainty of confidence level of the dented can, and x,y coordinates of a box around the dent.

Human Oversight Level

Human-Approved AI

Institutional Oversight Examples

  • •Video footage stays proprietary

Risk

Too many cans are identified as dented.

Output Modification Telemetry

  • Factual errors / hallucinations19%
  • Unfaithful outputs (policy, tone, or style misalignment)19%
  • Irrelevant information16%
  • Poor usability (unclear language)16%
  • Internal inconsistency16%
  • Missing or incomplete information14%

Transversal Metrics

Grouped by Fabric dimension.

Efficacy

Accuracy80 %

Efficiency

Modification Rate15 %
Modification Time1 min/output
Verification Time1 min
Rejection Rate8 %
Operational Friction4 hrs/wk
Implementation Overhead200 hrs
Governance Overhead100 hrs
Time to Launch12 wks

Value

Effort Reduction75 %

Risk

Reliability2 incident/month
Autonomy80 %