Core Technologies Overview

Ztrace Documentation

Ztrace leverages a sophisticated stack of AI and machine learning technologies to deliver industry-leading image intelligence capabilities.

Architecture Overview

Neural Network Architecture

Custom transformer-based models optimized for visual feature extraction and geospatial reasoning. Our models are trained on billions of geotagged images from around the world.

Multi-Modal Processing

Combines visual analysis with text recognition, satellite imagery matching, and contextual understanding for comprehensive geolocation.

Global Reference Database

Comprehensive dataset covering 190+ countries with continuous updates and validation. Over 50 billion reference points.

Model Versions

Model	Accuracy	Speed	Best For
geo-v3-turbo	91%	< 500ms	Real-time applications
geo-v3	94%	1-2s	General purpose
geo-v3-precision	97%	3-5s	Critical investigations

Processing Pipeline

precision-analysis.ts

TypeScript

1// Example: Custom model selection
2<span <span class=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>class</span>=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>const</span> result = <span <span class=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>class</span>=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>><span class=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>await</span></span> client.<span class="text-blue-400">analyze</span>({
3  image: <span class="text-emerald-<span class="text-orange-400">400</span>">'./evidence.<span class="text-cyan-300">jpg</span>'</span>,
4  options: {
5    model: <span class="text-emerald-<span class="text-orange-400">400</span>">'geo-v3-precision'</span>,  // <span class="text-yellow-300">Use</span> highest accuracy model
6    confidence_threshold: <span class="text-orange-400">0.<span class="text-cyan-300">9</span></span>,
7    include_features: <span <span class=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>class</span>=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>true</span>,
8    include_landmarks: <span <span class=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>class</span>=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>true</span>,
9    include_satellite: <span <span class=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>class</span>=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>true</span>,    // <span class="text-yellow-300">Cross</span>-reference <span <span class=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>class</span>=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>><span class=<span class="text-emerald-<span class="text-orange-400">400</span>">"text-purple-<span class="text-orange-400">400</span> font-medium"</span>>with</span></span> satellite data
10    max_results: <span class="text-orange-400">3</span>              // <span class="text-yellow-300">Return</span> top <span class="text-orange-400">3</span> location candidates
11  }
12});
13 
14// Access detailed results
15console.<span class="text-blue-400">log</span>(<span class="text-emerald-<span class="text-orange-400">400</span>">'<span class="text-yellow-300">Primary</span> <span class="text-yellow-300">Location</span>:'</span>, result.<span class="text-cyan-300">location</span>);
16console.<span class="text-blue-400">log</span>(<span class="text-emerald-<span class="text-orange-400">400</span>">'<span class="text-yellow-300">Confidence</span> <span class="text-yellow-300">Breakdown</span>:'</span>, result.<span class="text-cyan-300">confidence_breakdown</span>);
17console.<span class="text-blue-400">log</span>(<span class="text-emerald-<span class="text-orange-400">400</span>">'<span class="text-yellow-300">Alternative</span> <span class="text-yellow-300">Locations</span>:'</span>, result.<span class="text-cyan-300">alternatives</span>);

Last updated December 2025

Was this helpful?