Skip to main content

​
Overview

The MatchResult model represents an individual match result from a face comparison operation. It contains the person’s name and a probability score indicating how closely their face matches the target face.

​
Rust Definition

#[derive(Serialize)]
pub struct MatchResult {
    pub name: String,
    pub probability: f64,
}

​
TypeScript Definition

interface MatchResult {
  name: string;
  probability: number;
}

​
Fields

​
name
string
required
The name of the person from the original comparison request. This corresponds to the name field in the Person object that was compared.
​
probability
number
required
A floating-point value between 0.0 and 1.0 representing the probability that this person matches the target face. Higher values indicate stronger matches. A value of 1.0 would represent a perfect match, while 0.0 indicates no similarity.

​
Example JSON

{
  "name": "Jane Smith",
  "probability": 0.9823
}

​
Array Example

Match results are typically returned as an array in CompareResponse: 
[
  {
    "name": "Jane Smith",
    "probability": 0.9823
  },
  {
    "name": "Alice Johnson",
    "probability": 0.7456
  },
  {
    "name": "Bob Williams",
    "probability": 0.3219
  },
  {
    "name": "Charlie Brown",
    "probability": 0.1847
  }
]

​
Understanding Probability Scores

​
Score Ranges

The probability field represents the confidence level of the match:
RangeConfidenceInterpretation
0.9 - 1.0Very HighStrong match - very likely the same person
0.7 - 0.9HighGood match - likely the same person
0.5 - 0.7ModerateSome similarity - uncertain match
0.3 - 0.5LowWeak similarity - probably different people
0.0 - 0.3Very LowLittle to no similarity - likely different people

​
Factors Affecting Probability

Several factors can influence the probability score:
  • Image Quality: Higher resolution and clearer images produce more accurate scores
  • Lighting Conditions: Consistent lighting between target and comparison images improves accuracy
  • Face Angle: Frontal faces are easier to match than profile or angled shots
  • Age Difference: Significant time differences between photos may lower probability scores
  • Facial Expressions: Neutral expressions typically match better than extreme expressions
  • Obstructions: Glasses, masks, or other face coverings can affect matching accuracy
For different use cases, consider these threshold recommendations:
  • High Security Applications (e.g., access control): Use a threshold of 0.85 or higher
  • Standard Identification (e.g., photo tagging): Use a threshold of 0.70 to 0.80
  • Similarity Suggestions (e.g., “people you might know”): Use a threshold of 0.50 to 0.70
  • Research/Analytics: Adjust based on your specific accuracy requirements

​
Usage Example

Processing match results in your application: 
const response = await fetch('/compare', {
  method: 'POST',
  body: JSON.stringify(compareRequest)
});

const data: CompareResponse = await response.json();

// Find the best match
const bestMatch = data.matches.reduce((prev, current) => 
  current.probability > prev.probability ? current : prev
);

// Filter high confidence matches (threshold: 0.7)
const highConfidenceMatches = data.matches.filter(
  match => match.probability >= 0.7
);

if (bestMatch.probability >= 0.85) {
  console.log(`High confidence match found: ${bestMatch.name}`);
} else if (bestMatch.probability >= 0.7) {
  console.log(`Probable match: ${bestMatch.name}`);
} else {
  console.log('No confident matches found');
}
  • CompareResponse - Contains an array of MatchResult objects
  • Person - The source person model that generated this match result
  • CompareRequest - The request that initiates the comparison

Build docs developers (and LLMs) love

Get started for freeTalk to us