Video Intelligence for Mining

Ore Loss Detection & Production Optimization for Mining Operators

Lepei.pro brings AI video analytics to the mining industry helping operators identify ore losses, optimize processes, and save up to $200,000+ per year without upgrading equipment.

Project Summary:

Lepei.pro expanded into the mining sector. A partner from Australia initiated the pilot. Through several interviews, we uncovered how the same capabilities—video analytics, machine learning, computer vision—perfectly map to mining's core inefficiencies:

Unaccounted ore losses

Non-optimized haul truck routes

Quality fluctuations in processing

Financial leakage during crushing, screening, and flotation stages.

We created a suite of edge-deployable AI tools that operate on-site (no internet required), analyze video feeds, and deliver actionable results in just 48 hours.

Faced Challenge:

Architectural blueprint of a building with intricate piping and structural details

1-Ore Loss Detection

Problem:
Mining sites often lose valuable ore through waste rock misclassification or underloading. These losses go unnoticed and unaccounted.

Impact:
Up to 3% of ore is wasted—equivalent to $200,000+ annually for mid-sized mines.

2-Process Deviations in Crushing and Flotation

Problem:
Manual inspection can't catch minor process drifts or defective outputs in real-time.

Impact:
Inconsistent quality and lower yields, costing $100k+ per year in reprocessing and missed output.

3-Non-optimal Logistics and Loading Operations

Problem:
Inefficient truck routing, underused loaders, and material mix-ups during stockpile management.

Impact:
Wasted fuel, equipment overuse, and unpredictable throughput.

Lepei’s AI-Powered Solution:

AI Loss Detector

How It Works:
Analyzes camera feeds at extraction points, conveyors, or processing areas.
Detects waste rock, underextracted ore, or misclassified materials
Flags loss zones with visual heatmaps and estimated financial impact.

Technology Stack:
Fast3R: Accelerated 3D reconstruction from up to 1500 frames.
SpatialLM: Scene interpretation and object classification.
Edge deployment: Operates directly on mining site hardware.

Benefits:
Recovers 1–3% of ore typically lost.
Adds $100,000–$500,000 in yearly value without changing infrastructure.

AI Production Inspector

How It Works:
Monitors video from processing lines.
Detects anomalies, predicts failures or drifts in production.
Delivers a performance report within 48 hours.

Technology Stack:
Agentic RAG: Auto-refines analysis using a self-questioning agent.
Self-adaptive context engine: Learns mine-specific conditions—ore type, lighting, equipment behavior.

Benefits:
Reduces process-related losses by 20%.
Cuts reprocessing costs and boosts yield predictability.

Implementation: Step-by-Step Roadmap

We provide a 4-week structured pilot program that lets client test and validate the AI workflow before a full-scale deployment.

Phase 1: Pilot Launch

Upload site videos (drone, CCTV, loader cam, etc.)

Define target KPIs (loss %, defect rate, throughput)

Phase 2: AI Detection + Insight Report

Run AI models on supplied footage

Deliver report: heatmaps, before/after comparisons, savings potential

Phase 3: Validation & Scale

Validate results with internal ops team

Optional deployment to edge devices

Define long-term integration plan

Business Benefits:

Pilot found 3% ore losses this $200k/year saved at Australian mine.

Typical deployment ROI: <2-3 month

No need to change hardware or retrain teams

Builds digital transparency across sites

For more information on the technologies used, please follow the link

Ready to see what you're missing?

Get In Touch

Video Intelligence for Mining

Ore Loss Detection & Production Optimization for Mining Operators

Lepei.pro brings AI video analytics to the mining industry helping operators identify ore losses, optimize processes, and save up to $200,000+ per year without upgrading equipment.

Project Summary:

Project Summary:

Lepei.pro expanded into the mining sector. A partner from Australia initiated the pilot. Through several interviews, we uncovered how the same capabilities—video analytics, machine learning, computer vision—perfectly map to mining's core inefficiencies:

Unaccounted ore losses

Non-optimized haul truck routes

Quality fluctuations in processing

Financial leakage during crushing, screening, and flotation stages.

We created a suite of edge-deployable AI tools that operate on-site (no internet required), analyze video feeds, and deliver actionable results in just 48 hours.

Faced Challenge:

1-Ore Loss Detection

Problem:Mining sites often lose valuable ore through waste rock misclassification or underloading. These losses go unnoticed and unaccounted.

Impact: Up to 3% of ore is wasted—equivalent to $200,000+ annually for mid-sized mines.

2-Process Deviations in Crushing and Flotation

Problem: Manual inspection can't catch minor process drifts or defective outputs in real-time.

Impact: Inconsistent quality and lower yields, costing $100k+ per year in reprocessing and missed output.

3-Non-optimal Logistics and Loading Operations

Problem: Inefficient truck routing, underused loaders, and material mix-ups during stockpile management.

Impact: Wasted fuel, equipment overuse, and unpredictable throughput.

Lepei’s AI-Powered Solution:

AI Loss Detector

How It Works: Analyzes camera feeds at extraction points, conveyors, or processing areas. Detects waste rock, underextracted ore, or misclassified materials Flags loss zones with visual heatmaps and estimated financial impact.

Technology Stack: Fast3R: Accelerated 3D reconstruction from up to 1500 frames. SpatialLM: Scene interpretation and object classification. Edge deployment: Operates directly on mining site hardware.

Benefits: Recovers 1–3% of ore typically lost. Adds $100,000–$500,000 in yearly value without changing infrastructure.

AI Production Inspector

How It Works: Monitors video from processing lines. Detects anomalies, predicts failures or drifts in production. Delivers a performance report within 48 hours.

Technology Stack: Agentic RAG: Auto-refines analysis using a self-questioning agent. Self-adaptive context engine: Learns mine-specific conditions—ore type, lighting, equipment behavior.

Benefits: Reduces process-related losses by 20%. Cuts reprocessing costs and boosts yield predictability.

Implementation: Step-by-Step Roadmap

Implementation: Step-by-Step Roadmap

We provide a 4-week structured pilot program that lets client test and validate the AI workflow before a full-scale deployment.

Phase 1: Pilot Launch

Upload site videos (drone, CCTV, loader cam, etc.)

Define target KPIs (loss %, defect rate, throughput)

Phase 2: AI Detection + Insight Report

Run AI models on supplied footage

Deliver report: heatmaps, before/after comparisons, savings potential

Phase 3: Validation & Scale

Validate results with internal ops team

Optional deployment to edge devices

Define long-term integration plan

Business Benefits:

Business Benefits:

Pilot found 3% ore losses this $200k/year saved at Australian mine.

Typical deployment ROI: <2-3 month

No need to change hardware or retrain teams

Builds digital transparency across sites

For more information on the technologies used, please follow the link

Problem:
Mining sites often lose valuable ore through waste rock misclassification or underloading. These losses go unnoticed and unaccounted.

Impact:
Up to 3% of ore is wasted—equivalent to $200,000+ annually for mid-sized mines.

Problem:
Manual inspection can't catch minor process drifts or defective outputs in real-time.

Impact:
Inconsistent quality and lower yields, costing $100k+ per year in reprocessing and missed output.

Problem:
Inefficient truck routing, underused loaders, and material mix-ups during stockpile management.

Impact:
Wasted fuel, equipment overuse, and unpredictable throughput.

How It Works:
Analyzes camera feeds at extraction points, conveyors, or processing areas.
Detects waste rock, underextracted ore, or misclassified materials
Flags loss zones with visual heatmaps and estimated financial impact.

Technology Stack:
Fast3R: Accelerated 3D reconstruction from up to 1500 frames.
SpatialLM: Scene interpretation and object classification.
Edge deployment: Operates directly on mining site hardware.

Benefits:
Recovers 1–3% of ore typically lost.
Adds $100,000–$500,000 in yearly value without changing infrastructure.

How It Works:
Monitors video from processing lines.
Detects anomalies, predicts failures or drifts in production.
Delivers a performance report within 48 hours.

Technology Stack:
Agentic RAG: Auto-refines analysis using a self-questioning agent.
Self-adaptive context engine: Learns mine-specific conditions—ore type, lighting, equipment behavior.

Benefits:
Reduces process-related losses by 20%.
Cuts reprocessing costs and boosts yield predictability.