# Luminar Network (SN87) Decentralized Video Surveillance Agents & The Agentic Economy Luminar Network (Subnet 87) is a decentralized vision intelligence infrastructure providing production-grade perception and forensic event reconstruction for mission-critical security and regulated enterprise environments. By integrating **Bittensor’s incentive framework** with a **hardware-agnostic Video Management System (VMS)**, Luminar decouples physical cameras from intelligence execution. This allows for scalable deployment across any multi-vendor network without vendor lock-in. The network functions as a competitive marketplace where miners deliver: * **Multi-Object Tracking:** Persistent identity across complex scenes. * **Anomaly Detection:** Real-time identification of security threats. * **Localization Adaptation:** Tailoring vision models to regional contexts. * **Retroactive Timeline Reconstruction:** Forensic causal event mapping. # Introduction Video surveillance has become one of the largest sources of real-world data in modern infrastructure. Cities, enterprises, financial institutions, and public agencies collectively operate hundreds of millions of cameras, generating continuous streams of visual intelligence. ### The Current Challenge: Proprietary Silos Despite the massive scale of data, most surveillance deployments remain fundamentally limited by: * **Fragmented Ecosystems:** Closed hardware stacks that prevent cross-platform integration. * **Vendor Lock-in:** Software platforms that restrict interoperability and slow down innovation. * **Centralized Control:** Proprietary architectures that centralize control of data and analytics, restricting experimentation. ### A New Class of Intelligence Advances in decentralized compute networks and agentic AI systems are enabling a new class of intelligence infrastructure that is modular, composable, and economically aligned. **Bittensor** introduces a permissionless marketplace for specialized machine intelligence. In this ecosystem, independent agents compete and collaborate to solve narrowly defined tasks with measurable performance incentives. This creates a unique opportunity to rethink how video intelligence is produced, distributed, and scaled. *** ### The Luminar Solution: A Universal Intelligence Layer Luminar Network addresses the gaps in the current security landscape by acting as a **Universal Intelligence Layer** for vision systems. It decouples hardware (CCTV and sensors) from intelligence (AI agents), enabling any compatible video source to access a continuously evolving pool of specialized vision capabilities. #### Key Features * **Hardware Agnostic:** Access high-level intelligence without being locked into specific camera brands or device-specific software stacks. * **Crowdsourced Excellence:** Leverages the Bittensor network to solve high-value tasks like localized anomaly detection and retroactive timeline reconstruction. * **Agentic Vision Hub:** Continuously interprets live and historical streams, correlates multi-modal data, and reconstructs causal event narratives. #### Impact This architecture transforms surveillance infrastructure from **static recording systems** into **adaptive, intelligence-driven platforms** that improve over time through open competition and decentralized incentives. Luminar provides evidentiary integrity for clients ranging from government bodies to global financial institutions. # System Architecture Luminar Network is designed as a modular, layered architecture that decouples physical sensing infrastructure from decentralized intelligence execution. The system consists of two primary layers: 1. **The Application Layer:** Interfaces directly with operators and physical devices. 2. **The Subnet Layer:** Hosts decentralized agentic intelligence through Bittensor miners and validators. This separation enables the independent evolution of hardware integration, intelligence capabilities, and economic coordination, while preserving operational reliability and data sovereignty for end users. *** ### 2.1 Application Layer The Application Layer is anchored by **Luminar’s Video Management System (VMS)**, a hardware-agnostic orchestration platform designed to integrate seamlessly with heterogeneous camera networks and sensor environments. The VMS supports omnidirectional interoperability across major CCTV vendors and edge devices, enabling organizations to modernize intelligence capabilities without replacing existing infrastructure or accepting vendor lock-in. #### Operational Environment The platform is engineered for deployment in mission-critical environments where availability, auditability, and data integrity are non-negotiable. Typical deployments include: * Traffic management systems * Border and perimeter security installations * Industrial manufacturing facilities * Enterprise security operations centers #### Core Functions Within the overall system architecture, the VMS performs two core functions: * **Data Ingestion and Normalization:** The VMS acts as the primary entry point for real-time video streams and sensor telemetry into the Luminar subnet. It standardizes heterogeneous inputs into a unified data pipeline, enabling consistent downstream processing regardless of source hardware or protocol. * **Intelligence Delivery and Control Plane:** The VMS serves as the distribution layer through which AI-generated insights, alerts, and reconstructed event narratives are delivered back to operators and downstream systems. This bidirectional interface allows organizations to maintain full operational control over data routing, retention policies, and compliance boundaries. > By separating device management from intelligence execution, the Application Layer transforms traditional VMS deployments from static recording systems into adaptive intelligence gateways. *** ### 2.2 Subnet Layer (Miners and Validators) The Subnet Layer operates on Bittensor as a decentralized marketplace for specialized vision intelligence. Rather than monolithic processing pipelines, miners function as **autonomous computational agents** optimized for narrowly scoped, high-value tasks. #### Competitive Ecosystem * **Miners:** Compete to deliver measurable performance improvements within specialization domains, with rewards allocated via Bittensor’s incentive mechanism. * **Validators:** Continuously assess miner outputs using objective evaluation metrics, enforcing alignment between economic incentives and technical quality. This feedback loop creates a self-optimizing ecosystem where new capabilities emerge organically and high-performing agents are scaled by market demand. #### Foundational Agentic Capabilities The initial benchmark focuses on evaluating the following capabilities: 1. **Localization and Context Adaptation:** Fine-tuning perception models for geographic, regulatory, and cultural environments (e.g., regional license plates, signage conventions, uniforms, and behavioral norms). 2. **Semantic Data Labeling and Curation:** Structuring raw video streams into high-quality annotated datasets that improve downstream training efficiency, traceability, and auditability. 3. **Retroactive Timeline Reconstruction:** Correlating fragmented multi-camera footage to reconstruct entity trajectories, causal event sequences, and historical movement patterns across time and space. Together, these capabilities establish the foundational intelligence primitives required for scalable forensic analysis, anomaly detection, and long-horizon situational awareness across decentralized infrastructure. # Agentic Competition and Benchmarking A core design principle of Luminar Network is that intelligence quality must emerge from open competition rather than static model deployment. Instead of relying on centrally trained monolithic models, Luminar leverages Bittensor’s incentive mechanism to continuously benchmark, rank, and economically reward specialized agentic behaviors under realistic operating conditions. #### The Role of Benchmarks Benchmarks serve three critical functions within the network: * **Objective Performance Measurement:** Establishing standardized, reproducible metrics for evaluating agent quality under production-like workloads. * **Economic Signal Generation:** Translating technical performance into incentive-aligned rewards that drive miner optimization and specialization. * **Capability Evolution:** Enabling rapid iteration and emergence of new intelligence primitives without centralized model governance. This framework ensures that only agents demonstrating operational reliability, temporal consistency, and latency-aware performance are promoted within the subnet and exposed to downstream applications. *** ### 3.1 The Benchmark: Luminar Multi-Object Tracking and Anomaly Benchmark (L-MOT) To operationalize agentic competition, we introduce the **Luminar Multi-Object Tracking and Anomaly Benchmark (L-MOT)**. L-MOT evaluates a miner’s ability to maintain persistent situational awareness across continuous video streams while simultaneously detecting security-relevant behavioral anomalies. #### Testing Environment Miners are provided with curated test video segments that emulate real-world surveillance challenges, including: * Occlusions and variable lighting. * Camera motion and dense multi-object scenes. Each miner must execute two concurrent tasks under strict latency constraints: 1. **Multi-Object Tracking:** Maintain consistent identities for dynamic entities (e.g., vehicles, individuals, assets) across frames, camera transitions, and partial occlusions. 2. **Event and Anomaly Recognition:** Detect and classify predefined anomalies (e.g., unattended objects, perimeter breaches, abnormal dwell time) with high precision. *** #### 3.1.1 Evaluation Metrics and Scoring Validators assess miner outputs using the **Higher Order Tracking Accuracy (HOTA)** metric, augmented with a latency-sensitive penalty function to discourage slow solutions. $$Score\_{agent} = HOTA \times (1 - LatencyPenalty)$$ **Understanding HOTA** HOTA jointly captures two complementary dimensions of tracking quality: * **Detection Accuracy (DetA):** Measures the agent’s ability to correctly identify and localize objects present in each frame. * **Association Accuracy (AssA):** Measures the agent’s ability to preserve identity consistency across time, ensuring an entity is correctly associated with its later appearances despite scale variations or viewpoint changes. #### Operational Standards The latency penalty incorporates end-to-end inference delay relative to real-time thresholds. By coupling spatial accuracy, temporal consistency, and execution efficiency into a single economic signal, L-MOT incentivizes miners to optimize for **production-grade intelligence** rather than academic performance. This ensures that the intelligence surfaced to the Application Layer meets the reliability, responsiveness, and evidentiary standards required for operational security and forensic reconstruction. # Incentive Mechanism The incentive mechanism constitutes the cryptoeconomic control layer of Subnet 87, governing how decentralized intelligence is evaluated, priced, and continuously improved. Rather than relying on trusted intermediaries, the network embeds performance accountability directly into its economic structure through Bittensor’s native validation and reward distribution framework. This converts technical signals—accuracy, temporal consistency, and latency—into direct financial incentives, creating a closed feedback loop between engineering and economic outcomes. *** ### 4.1 Validation of Data Labeling For tasks such as object detection, semantic segmentation, or activity classification, validators must ensure the accuracy and consistency of labeled outputs. Subnet 87 employs a comprehensive benchmark evaluation methodology. Validators maintain a curated benchmark dataset $D\_{benchmark}$ consisting of pre-labeled ground truth annotations across diverse scenarios. When evaluating a miner’s labeling capability, the entire benchmark is presented: $$D\_{evaluation} = D\_{benchmark}$$ The miner’s score $S\_m$ is calculated by measuring prediction accuracy using a task-appropriate loss function $L$: $$S\_m = e^{-\alpha \cdot L(Label\_{miner}, Label\_{ground\_truth})}$$ * **$\alpha$:** A scaling parameter controlling the sensitivity of the exponential decay. * **Purpose:** This prevents miners from selectively optimizing for specific data subsets, as every sample in the benchmark contributes to the final score. *** ### 4.2 Validation of Localization Effective video intelligence requires models that understand geographic, cultural, and temporal context. The localization task evaluates a miner’s ability to adapt models to region-specific characteristics (e.g., Tokyo vs. Berlin). Validators present miners with geographically and temporally tagged samples $V\_{region,time}$. Let $C$ represent the set of contextual features: $$C = { \text{traffic signs, vehicle types, behavioral norms, traffic rules} }$$ The localization score $S\_l$ is computed as: $$S\_l = \frac{1}{|C|} \sum\_{c \in C} \text{Accuracy}(Prediction\_{miner}^{c}, GroundTruth^{c}\_{r,t})$$ This metric ensures miners can distinguish between regional infrastructure and social norms, incentivizing geographically adaptive models over monolithic systems. *** ### 4.3 Validation of Retroactive Timeline Construction This task requires miners to reconstruct a subject’s movement by stitching together fragmented footage from multiple camera sources. #### Asymmetric Verification The methodology uses **asymmetric verification**, where computationally expensive backward reconstruction serves as the ground truth: 1. **Backward Flow (Ground Truth):** Validators work backward from a known endpoint to trace a path. This is highly accurate but too slow for real-time use. 2. **Forward Flow (Miner Task):** Miners must predict the trajectory moving forward in real-time based on appearance and spatiotemporal reasoning. The reward $R$ is calculated using **cosine similarity** between the miner's path vector $\vec{P}*{miner}$ and the validator's ground truth $\vec{P}*{validator}$: $$R = \frac{\vec{P}*{miner} \cdot \vec{P}*{validator}}{|\vec{P}*{miner}| |\vec{P}*{validator}|}$$ *** ### 4.4 Overall Validation All benchmark tasks within an evaluation epoch are jointly validated as a single competitive batch. Miner scores across localization, labeling, and timeline reconstruction are aggregated into a unified performance metric. #### Winner-Takes-All Model At the end of each cycle, the highest-performing miner receives **100% of the emissions** allocated for miners for that epoch. This model enforces: * Strong competitive pressure. * Accelerated model iteration. * Guaranteed economic rewards for only the most reliable agents. # Tokenomics & Emission Control The **Proof-of-Value** emission model is explicitly linked to application-layer deployments of Guardian and Flow Pilots. Emissions unlock only upon verified revenue from VMS contracts, enforcing strict coupling between decentralized intelligence and real-world adoption. *** ### 5.1 The Burn Mechanism Luminar implements a radical "Proof of Value" economic model to prevent token inflation without utility. #### Default State $$E\_{miner} = 0 \quad (100% \text{ Burn})$$ #### Revenue Triggered State Emissions are unlocked only when the application layer generates verified revenue (fiat or crypto flow from VMS contracts). $$E\_{miner} = \min(E\_{cap}, 0.4 \times R\_{application})$$ **Variables:** * **$E\_{miner}$:** The emission released to miners. * **$R\_{application}$:** The revenue generated by the Luminar VMS. * **$0.4$:** Represents the 40% revenue-to-emission ratio. *** ### 5.2 Emission Growth Strategy (Targeting 0.5% Network Emission) The subnet’s growth strategy is designed to sustainably increase its emission share by strengthening economic credibility and validator consensus weight. Rather than relying on speculative signaling, the strategy prioritizes measurable capital alignment and revenue-backed reinforcement. The following levers are employed to raise the subnet’s emission floor: #### 1. Protocol-Aligned Buyback and Burn A defined portion of operating surplus generated from VMS deployments is allocated toward periodic **SN87 alpha token buyback and burn** operations. This mechanism: * Reinforces long-term token scarcity. * Aligns commercial success with network value accrual. * Signals sustained operational commitment to the ecosystem. #### 2. Validator Stake Accretion Structured outreach and ecosystem engagement initiatives are used to attract long-term token holders to delegate stake to Luminar-operated validators. * **Consensus Weight:** Increased delegated stake strengthens validator consensus. * **Stability:** Improves emission stability and enhances subnet visibility within the broader Bittensor network. > This dual approach couples real economic throughput with network-level participation incentives, ensuring growth is grounded in operational performance rather than short-term capital inflows. # Business and Roadmap ### 6.1 Go-to-Market Strategy Luminar’s commercialization strategy follows a phased market entry model, prioritizing high-value, compliance-driven environments where reliability, auditability, and evidentiary integrity are mandatory. #### Target Verticals * **Government and Public Sector Adoption:** Initial deployments target municipal and public safety infrastructure, including traffic monitoring systems, law enforcement surveillance, and urban security operations. Market entry is driven by domain expertise and operational credibility to accelerate procurement cycles. * **Private Sector Expansion:** Following public-sector validation, the platform expands into regulated enterprise environments: * **Banking Infrastructure:** ATM monitoring and fraud detection. * **Commercial Complexes:** Footfall analytics and incident detection. * **Industrial Facilities:** Safety compliance and perimeter monitoring. > This sequencing reduces go-to-market risk, validates system robustness under demanding conditions, and establishes reference deployments before horizontal scaling. *** ### 6.2 Product Roadmap and Capability Expansion Platform development proceeds through staged capability activation aligned with operational maturity and market demand. #### Phase 1: Agentic Competition Activation Launch of the **L-MOT benchmark** and validator scoring infrastructure to establish baseline intelligence quality, miner specialization dynamics, and economic feedback loops. #### Phase 2: Omnidirectional VMS Deployment Production rollout of the hardware-agnostic **VMS layer**, expanded integration partnerships, and field deployments across heterogeneous camera environments. #### Phase 3: Advanced Forensic Intelligence Full integration of **retroactive timeline reconstruction**, multi-camera causal reasoning, and evidentiary-grade analytics pipelines for law enforcement and regulated enterprise use cases. #### Phase 4: Autonomous Security Agents and Predictive Intelligence Introduction of **autonomous agent orchestration**, where multiple specialized miners collaborate for: * Proactive threat assessment and behavioral risk forecasting. * Early-warning detection (e.g., crowd-risk escalation alerts). * Anomaly trend forecasting and adaptive perimeter reinforcement. In this phase, intelligence outputs transition from isolated alerts to continuously updated **probabilistic threat models**, positioning Luminar as a decentralized, self-optimizing security intelligence fabric. *** *Each phase incrementally expands the system surface area while maintaining operational stability and measurable performance guarantees.* # Conclusion Luminar Network functions as a production-grade intelligence substrate that bridges decentralized computation with the operational rigor of real-world security environments. By combining a **revenue-aligned emission strategy**, **transparent agentic benchmarking**, and **decentralized cryptoeconomic enforcement**, the network ensures that artificial intelligence capabilities translate into verifiable, actionable intelligence at scale. This architecture enables: * **Continuous Evolution:** Performance improves without centralized dependency. * **Operational Rigor:** Preserves reliability and auditability. * **Mission-Critical Readiness:** Meets the governance properties required for deployments across public and private infrastructure.