Skip to content
/How Image Verification Works

How Image Verification Works

Camera Technologyintermediate3mo ago
1 view

A walkthrough of MaviCats' three-tier image verification pipeline — from content safety to sensor fingerprinting.

Overview

Every image uploaded to MaviCats passes through a three-tier verification pipeline. This system confirms content safety, verifies camera authenticity, and provides trust signals for the community.

Tier 1: Safety Scan

The first tier runs immediately after upload and checks for content safety:

  • Uses a neural network classifier (ResNet-50) to analyze image content
  • Produces a confidence score for content appropriateness
  • Images scoring below the safety threshold are flagged for human review
  • Results appear as a green "Verified" or yellow "Under Review" badge

Typical processing time: 100–500ms

Tier 2: Camera Verification

The second tier reads EXIF metadata to identify the camera:

  • Extracts Make and Model tags from the JPEG header
  • Fuzzy-matches the model name against the MaviCats Gearbase database
  • Validates that the image resolution matches the camera's known output specs
  • If the user selected a camera when posting, checks for consistency

Possible verdicts:

  • Camera Verified (green) — EXIF matches the claimed camera and resolution is correct
  • Camera Mismatch (amber) — EXIF data conflicts with the user's camera selection
  • Unverified — No EXIF data found or no camera match

Tier 3: Sensor Fingerprint

The third tier performs Sensor Pattern Noise (SPN) analysis:

  • Extracts the noise residual from the uploaded image using wavelet denoising
  • Compares it against stored reference signatures for the matched camera model
  • Computes Normalized Cross-Correlation (NCC) between the image noise and reference signature
  • High correlation confirms the image was taken by the identified camera type

Possible verdicts:

  • Sensor Verified (cyan) — Strong NCC correlation with reference signature
  • Sensor Plausible (blue) — Moderate correlation, consistent with the camera model
  • Unmatched — Correlation too low or no reference signature exists
  • Insufficient — Reference signature not available for this camera

What you see

After uploading, you'll see badges appear on your images as each tier completes:

  1. A scan status badge (Scanning → Verified/Under Review)
  2. Camera verification badge (if EXIF data is found)
  3. Sensor fingerprint badge (if a reference signature exists)

Post owners can expand the "Image Verification Pipeline" panel on their post's detail page to see full results from all three tiers.

Why this matters

For a community built around vintage cameras, authenticity is core to the experience. The verification pipeline helps:

  • Confirm Mavica origin — green badges signal genuine lo-fi photos
  • Detect reposts — images from modern cameras tagged as Mavica will show mismatches
  • Build trust — verified sellers and collectors can prove their photos are authentic
  • Preserve the aesthetic — the community can celebrate real imperfections over digital filters