The State of Play

3D sensing and reconstruction has reached production deployment at forward-leaning organisations, but the field remains structurally bifurcated and most organisations have not yet started. Photogrammetry pipelines have commoditised into profitable surveying infrastructure across construction, mining, and land management, with documented 96% positive ROI among active users. LiDAR anchors autonomous vehicle perception and critical infrastructure mapping, with over 800,000 vehicles now equipped. Yet these two modalities serve fundamentally different markets: photogrammetry is accessible and cost-effective but limited to roughly 20mm accuracy; LiDAR delivers sub-2mm precision but carries persistent cost, weather sensitivity, and vendor consolidation risks. The tier-defining tension is clear. Algorithms, tooling, and ecosystem partnerships are production-ready. Sensor reliability and environmental robustness are not -- constraining scaled adoption in safety-critical domains and keeping this practice at the vanguard rather than the mainstream.

Automotive LiDAR remains at inflection: 15+ brands with production vehicles on market, robotaxi fleets from Baidu, Waymo, and Cruise at scale, with market trajectory to $4.5B by 2028 (55% CAGR). Photogrammetry deployment economics have solidified with field evidence of production parity. April 2026 case study: professional surveyor field validation confirmed iPhone 17 Pro + RTK achieves 2cm–9cm accuracy matching professional total stations with 30x labor reduction, demonstrating consumer-device viability for commercial surveying workflows. Commercial construction deployment (Henderson, Nevada) verified DJI RTK accuracy (±0.05 ft horizontal) with $18k cost avoidance outcome. Algorithmic maturity is progressing: transformer-based point cloud segmentation research (Point Transformer V3, Swin3D) demonstrates +18% mIoU improvement with practical pre-labeling workflows in construction domains. Ecosystem indicators: ISPRS benchmarks standardizing on WHU-TLS (1.74B points, 11 environments), Claru AI commercial dataset product (60K+ annotated urban scans across 40+ cities) addressing industry training-data gaps. Hyundai's real-world deployment across 22,000 km on-road driving achieved 95%+ LiDAR contamination classification accuracy—technical mitigation for Level 4 autonomous vehicle barriers.

Yet critical robustness barriers remain unresolved. Urban environment analysis (April 2026) documents active deployment testing against real constraints: occlusions, small-object detection failures, weather-induced point loss (59% detection reduction in fog). Wet-road comparative testing confirms fundamental limitation: current 3D sensing cannot measure road friction coefficient, doubling stopping distance in rainy conditions—requiring hardware solutions (1550nm wavelength, event cameras) rather than algorithmic fixes. Multi-platform LiDAR study (urban tree inventory, 427,000+ trees) identifies remaining constraints: species classification unsolved, DBH uncertainty dominates, crown condition assessment requires manual inspection. Practitioner analysis documents persistent accuracy bifurcation: Matterport ±20mm versus terrestrial LiDAR ±1.9mm, constraining photogrammetry to documentation. LiDAR unit costs plateau at $1k, prohibitive for consumer deployments. The technology split hardens: photogrammetry scales profitably in surveying and civil engineering; LiDAR anchors autonomous mobility and infrastructure inspection—but scaling to mainstream adoption requires resolved weather robustness and cost accessibility neither platform has yet achieved.

June 2026 scan findings confirm market acceleration alongside persistent architectural constraints and emerging architectural competition. Ouster announced Rev8 native color LiDAR with doubled range/resolution and broad OEM adoption (Google, Volvo, Liebherr, Epiroc, Seegrid), integrating depth and color at hardware level and signaling production-maturity convergence in sensor design. RoboSense Q1 2026 results show structural inflection: robotic LiDAR shipments surged 1,458.8% year-over-year to 185,500 units (56% of total), exceeding ADAS for the first time and spanning robotic lawnmowers, autonomous delivery, humanoid robots, and cleaning systems—indicating adoption expanding beyond automotive into multi-sector robotics. Intel's Mobileye Drive announced production commitment: nine InnovizTwo LiDAR sensors per vehicle (360-degree coverage) scaling from 100 vehicles in 2027 to 17,000 over five years, representing 150,000+ unit opportunity in commercial L4 robotaxi platform. Production deployments confirm maturity: Hitachi Construction Machinery live digital twin platform with 3D laser scanning achieved ±20mm accuracy and 45ms latency on remote excavator teleoperation, outperforming competitor systems (150–300ms); Artec Jet SLAM LiDAR launched with multiple case studies (concert halls, mines, heritage sites, power lines, infrastructure) demonstrating GPS-independent 3D mapping in challenging environments; Gecko Robotics deployed Ouster Rev8 for infrastructure asset mapping and digital twins at scale across energy, O&G, and military inspections. Government procurement signals: Niantic Spatial's 3D reconstruction platform achieved DoW Tradewinds "Awardable" status and deployed across multiple U.S. military services for geospatial intelligence, demonstrating commercial maturity at leading-edge scale. Photogrammetry adoption on mega-projects: drone surveying documented as standard contract requirement on Saudi Arabia's NEOM ($500B), Qiddiya, and Diriyah projects, with RTK-equipped crews mapping 50–150 hectares/day at 1–3cm accuracy. Yet critical architectural alternatives are gaining traction: Applied Intuition's Self-Driving System achieved production deployment in North America, Europe, and Japan WITHOUT LiDAR or HD maps, relying instead on camera-radar fusion with onboard neural networks—demonstrating that cost and complexity barriers limit LiDAR adoption to specific use-cases rather than universal necessity. Robustness limitations persist: research identified critical solid-state LiDAR failure mode (internal-multipath glare creating phantom objects) requiring algorithmic mitigation; Aptiv documented specific warehouse deployment failures (reflective materials, transparent barriers, thin occluded objects causing false detections) driving multi-sensor fusion requirements; BitSensing analysis quantified adverse-weather degradation (rain/fog backscattering and attenuation) limiting perception reliability. Autonomous system evaluation confirmed 98% success under isolated disturbances versus 52% under combined adverse conditions (fog+rain+night+traffic), documenting domain-shift barriers and the gap between controlled testing and real-world operation.

• 2018: Photogrammetry deployed in archaeology, geology, and construction monitoring; commercial tools (Pix4DMapper, Agisoft) show cost-competitiveness with laser scanners. LiDAR precision improvements ongoing but accuracy claims questioned by industry experts; practical deployments required ground control points to achieve survey-grade results.
• 2019: Commercial LiDAR adoption expanded into autonomous vehicle navigation and smart city urban mapping; photogrammetry and LiDAR ecosystems matured with standardized software tools for classification and design workflows. Systematic barriers persisted: accuracy divergence between reported specifications and field results, software reliability issues (point cloud corruption in Trimble), and continued need for use-case-specific technology selection.
• 2020: Photogrammetry tooling matured (Pix4Dmatic GA), enabling large-scale image processing and production deployment across agriculture, disaster mapping, and cadastral surveying. LiDAR moved into production autonomous vehicles (Local Motors Olli, Ford Otosan heavy trucks) and active construction sites with quantified productivity benefits. Low-cost sensor market fragmented (Livox met specs, Ouster didn't), while reliability and security concerns (fault detection, spoofing attacks) emerged. Technology split: photogrammetry dominated cost-sensitive surveying; LiDAR became essential for autonomous mobility and high-precision infrastructure.
• 2021: Academic progress in real-time photogrammetry and drone-based LiDAR systems demonstrated sub-meter and decimetre-level accuracy; product ecosystem expanded with survey-grade mobile 3D scanning (Pix4D RTK). Vendor ecosystem matured with standardization initiatives (Cloud Optimized Point Cloud format for cloud-native LiDAR). Applied research validated LiDAR on construction sites (robotics integration for monitoring). Real-world deployments continued to surface accuracy challenges and Z-axis errors in production workflows, revealing practical implementation gaps despite theoretical advances.
• 2022-H1: LiDAR deployments advanced in autonomous vehicles (Roborace racing series using Velodyne Velarray H800). Photogrammetry vendor maturity signaled through independent certification (Pix4D viDoc RTK <5cm accuracy by Bureau Veritas). Academic research identified persistent challenges in 3D reconstruction—surface reconstruction from point clouds continued to struggle with multi-view alignment, missing surfaces, and outlier handling. Empirical studies quantified photogrammetric requirements (50-80+ points/m² for building detail). Mixed reality platform tooling matured (Microsoft HoloLens spatial mapping APIs formalized).
• 2022-H2: Production 3D sensing expanded across industrial, infrastructure, and autonomous vehicle sectors—AWS SageMaker integrated Velodyne LiDAR for enterprise perception datasets; photogrammetry deployments scaled (Pix4Dmatic processing 10,500-hectare surveying missions); LiDAR ecosystem began consolidation. However, Velodyne-Ouster merger amid combined $72M+ annual losses indicated hardware vendor profitability pressures despite growing sensor volumes. Surface reconstruction and Z-axis accuracy remained unresolved research challenges.
• 2023-H1: Photogrammetry tools expanded into forensic reconstruction (accident analysis), mobile spatial apps, and facility digitalization. LiDAR supplier base contracted from 65-70 (2018) to ~25 companies; Hesai scaled production (80k units in 2022, 1M/year factory capacity), while Ouster-Velodyne pivot to infrastructure signaled AV-only model unsustainability. Market forecasts projected 3D mapping growth to $26.9B by 2033 (16.42% CAGR). However, peer-reviewed research identified critical constraint: LiDAR performance degradation in adverse weather (dense fog, heavy rain severely impact detection), and surface reconstruction remained algorithmically unresolved. Autonomous vehicle LiDAR validated for map updates (case study with Velodyne VLS-128), but profitability and weather reliability barriers limited scaling.
• 2023-H2: Enterprise photogrammetry platforms solidified—Pix4D partnered with DroneGIS for enterprise cloud deployment (AWS, Azure, GCP) targeting construction and mining. Automotive LiDAR reached inflection point: Yole forecasts $4.48B market by 2028 (69% CAGR) with Hesai leading 67% share in robotaxis and 103k cumulative units deployed; Chinese OEMs and design wins from global OEMs (Volvo, Polestar, Nissan, Mercedes) confirmed volume trajectory. Infrastructure and surveying deployments continued—bridge damage detection reached 3mm accuracy with LiDAR, photogrammetry merged aerial/terrestrial data for complex surveying tasks. Analyst teardowns documented cost escalation and technology divergence: rotating-mirror and MEMS sensors dominated OEM pipelines while cost barriers ($500/unit for automotive LiDAR vs $300 mass-market targets) remained the profitability constraint.
• 2024-Q1: Photogrammetry and reconstruction algorithms advanced: CVPR 2024 published peer-reviewed building wireframe reconstruction from aerial LiDAR (36-42% accuracy improvement), and ML-based methods for mesh prediction validated across European cities (Zurich, Berlin, Tallinn). Deployment breadth expanded—DroneDeploy's 2024 survey of 1,447 global users showed 96% positive ROI with 25% saving >$100k annually; construction, mining, and infrastructure applications scaled (Magil Construction, Parsons orphan well detection, Nevada DOT 9,500-mile asset mapping, historic preservation pilots). Market research confirmed growth trajectory: 3D mapping/modeling market projected $32.2B by 2032 (16.9% CAGR from $7.8B in 2023). However, critical reliability barriers surfaced: peer-reviewed automotive journal analysis documented LiDAR failure modes in real-world driving, with environmental weak points and component failures limiting OEM confidence for production autonomous vehicles—highlighting that despite algorithmic and deployment advances, sensor reliability constraints persisted.
• 2024-Q2: Ecosystem accessibility and algorithmic maturity advanced in parallel. Pix4D and Emlid launched integrated mobile terrestrial scanning kit ($3,990) combining RTK positioning with photogrammetry for utilities, crash reconstruction, and volumetric measurement—broadening professional adoption. UAV LiDAR validation continued: peer-reviewed case study (Ohio State) demonstrated DJI Zenmuse L2 exceeded traditional total stations in vegetated terrain surveying, capturing 1.2M points/second through dense foliage. ML-based reconstruction progressed: NeRF enhanced with global Structure-from-Motion improved urban 3D reconstruction from UAV imagery; deep learning frameworks achieved >98% accuracy in automated Scan-to-BIM element reconstruction. Real-world deployments expanded—Namibia Flying Labs deployed Pix4Dmatic for sustainable urban development and informal settlement housing planning. Implementation challenges persisted: developer forum discussions documented photogrammetry failures (metadata injection, depth data errors) with Apple ObjectCaptureSession, revealing practical barriers in consumer-facing 3D reconstruction tools.
• 2024-Q3: Algorithmic advances continued with transformer-based 3D reconstruction achieving 60% improvement on sparse LiDAR data, and systematic research quantifying LiDAR SLAM accuracy evaluation methods for autonomous systems. Critical robustness evaluation documented pose estimation degradation under adverse conditions (odometry errors increasing from 0.5% to 80% under rain/fog/noise), reinforcing weather reliability barriers. Ecosystem integration advanced: FlytBase-Pix4D partnership automated drone-to-cloud workflows (Pix4D Flink), streamlining end-to-end data processing. Adoption metrics solidified: 800,000+ vehicles equipped with LiDAR-based ADAS (190% growth from 2021), with named deployments from Volvo and NIO. Market data confirmed: 3D mapping industry at $5.2B in 2024 with 16.3% CAGR through 2031, though cost and complexity barriers constrained entry for smaller organizations.
• 2024-Q4: Production-stage deployments expanded across utilities, transportation, and urban planning. LiDAR-based infrastructure monitoring scaled: utilities deployed conductor sag analysis and vegetation detection systems; Ouster BlueCity LiDAR rolled out to hundreds of U.S. intersections for pedestrian/traffic detection. Government-scale investment continued: USGS 3D Elevation Program committed $70.9M Federal and $22.9M non-Federal for nationwide LiDAR/IfSAR acquisition in FY24. Photogrammetry ecosystem matured with Pix4D releasing volume computation for mining/construction applications. Core research addressed persistent algorithmic challenges: peer-reviewed IEEE TPAMI benchmark documented learning-based reconstruction quality gains but highlighted superior robustness of traditional methods on real-world anomalies; automated change detection for urban digital twins achieved 100% accuracy on building identification tasks. Market consolidation continued among LiDAR vendors with Chinese companies (Hesai, RoboSense) capturing 80%+ global share. Technology split remained entrenched: photogrammetry dominated cost-sensitive surveying and infrastructure documentation, while LiDAR anchored autonomous vehicle perception and real-time 3D mapping—but weather sensitivity and vendor profitability constraints persisted as adoption barriers.
• 2025-Q1: Algorithmic and deployment maturity deepened despite emerging reliability constraints. Research advanced multi-view 3D reconstruction: IEEE TVCG survey synthesized methods for large-scale urban reconstruction with industrial requirements (scalability, human integration), signaling production readiness across domains; Fast3R demonstrated Transformer-based scaling (1000+ images in one forward pass). Real-world deployments continued expanding: photogrammetry case studies documented 70% efficiency gains in building facade inspection (Brazil) and cost-effective volumetric surveying (UK); production adoption confirmed across AEC and land management sectors. However, critical reliability barriers crystallized: systematic evaluation (SORBET framework) quantified LiDAR obstacle detection robustness failures (20%+ variance with minor input perturbations), and comprehensive accuracy assessments revealed affordable 3D depth sensors (Stereolabs, Intel RealSense) consistently underperformed manufacturer specifications in real-world conditions with edge distortions. These findings reinforced the structural tension: core reconstruction algorithms and ecosystem accessibility were maturing for production use, but sensor reliability and environmental robustness remained unresolved constraints for scaled deployment in safety-critical applications (autonomous vehicles, critical infrastructure).
• 2025-Q2: Photogrammetry deployment scaling continued with quantified efficiency gains in surveying (weeks-to-hours processing for cadastral work), while research validated UAV LiDAR accuracy parity with terrestrial laser scanning for forensic applications. Algorithmic progress advanced semantic segmentation on heterogeneous robotic platforms (22% mIoU improvements), extending adoption into field robotics. However, critical adoption barriers intensified: industry analysis documented LiDAR cost constraints ($1,000/unit) and performance limitations in adverse weather (stereo vision superior in fog/rain), competing alternative technologies gained traction. Automotive LiDAR market scaled with 1.6M units shipped (2024) and 120+ car models deployed, predominantly driven by Chinese suppliers (93% market share). Yet workflow implementation challenges persisted: practitioner analysis documented critical pitfalls in point cloud processing (segmentation failures, misclassification risks) that continue to constrain production adoption despite algorithmic maturity.
• 2025-Q3: Reconstruction algorithms continued maturation with NeRF+SfM hybrid methods scaling to ultra-small platforms (sub-100g UAVs) and vendor-backed open-source tooling (NVIDIA) claiming 30-minute photorealistic mesh generation. Real-world deployments expanded in land management and environmental monitoring: cadastral digitization (Colombia) reduced processing from weeks to 6 hours at 5cm accuracy; conservation mapping (Brazil) processed 2,800 images and 1.8B points for Atlantic Forest monitoring, enabling deforestation detection and AI training. Cost and scale trends accelerated: automotive LiDAR cost declined to ~$1k/unit (from $80k in 2018) with market projected at 53.6% CAGR through 2032. However, deployment success barriers persisted: critical assessment of LiDAR in airport occupancy monitoring documented durability failures, integration complexity, and high cost—indicating application-specific limitations and that scaling challenges remain despite technical maturity.
• 2025-Q4: Ecosystem expansion and algorithmic maturity deepened with major telco partnerships (NTT DOCOMO Pix4D service launch) and SOTA transformer-based reconstruction methods (PLANA3R: 0.07m translation error on benchmark datasets). Multi-modal 3D fusion advanced (WorldMirror), and part-level segmentation research matured (S2AM3D CVPR 2026 dataset with 100k+ samples). Consumer-grade LiDAR validation confirmed iPhone 14 Pro viability for environmental surveying. Photogrammetry platform modernization continued (Pix4Dmapper legacy cloud pipeline decommissioned, improved migration path). TLS market confirmed sustained growth (USD 4.82B in 2024, 8.25% CAGR through 2035). However, adoption barriers remained entrenched: LiDAR cost ($1k/unit) and reliability constraints persisted in complex environments (airport occupancy monitoring case study documented failures), weather sensitivity unresolved for autonomous systems, and profitability pressures continued consolidating vendor base toward Chinese suppliers (93%+ market share in automotive). Technology bifurcation hardened: photogrammetry profitable and scaling; LiDAR struggling with cost and reliability despite strong market projections.
• 2026-Jan: Production deployment acceleration with Velodyne's Intelligent Infrastructure Solution reaching scale in Helsinki (97% accuracy on traffic monitoring), and Pix4D completing ecosystem consolidation via PIX4Dmatic 2.0 GA. Algorithmic advances continued on point cloud segmentation (transformer-based methods improving mIoU by 3.1% on automotive datasets, industrial research achieving 21.7% tail-class improvement). Critical practitioner analysis documented persistent accuracy gaps (Matterport ±20mm vs. LiDAR ±1.9mm), reinforcing technology trade-offs and segmentation by application. Automotive LiDAR market projections strengthened (31.3% CAGR to 2032, $6.46B), but production costs and weather sensitivity remained unresolved deployment barriers.
• 2026-Apr: Patent analysis confirmed 41,939 active LiDAR patents with 15+ automotive brands now shipping production vehicles and robotaxi fleets from Baidu, Waymo, and Cruise at scale; market trajectory to $4.5B by 2028 (55% CAGR) reinforced. Photogrammetry continued proving production ROI: Arizona and Nevada civil engineering firms documented DJI RTK deployments achieving 0.04-foot and 1.8cm RMSE accuracy, eliminating $14k ground surveys. Critical assessment hardened weather-robustness barriers: comparative wet-road AEB testing confirmed LiDAR cannot measure road friction coefficient, doubling stopping distance in rain — a fundamental gap requiring new hardware approaches (1550nm wavelength, event cameras) rather than software fixes.
• 2026-May: Production deployment evidence solidified with consumer-device viability confirmed: professional surveyor field study (iPhone 17 Pro + RTK) achieved 2cm–9cm accuracy parity with total stations at 30x labor reduction. Multi-platform operational study (427,000+ urban trees) demonstrated scaling challenges: species classification and crown condition assessment remain manual bottlenecks despite technical maturity of TLS/MLS/ALS platforms. Hyundai real-world validation (22,000 km on-road) achieved 95%+ contamination classification—practical sensor robustness mitigation for L4 autonomy. Urban deployment analysis confirmed active testing against robustness barriers: fog reduces detection 59%, occlusions and small-object failures persist. Algorithmic maturity advanced: transformer-based segmentation (+18% mIoU) enables construction-domain pre-labeling workflows. Commercial dataset products (60K+ annotated urban scans) signal ecosystem recognition of training-data gap. Hardware frontier advanced: MIT silicon-photonics OPA breakthrough enables compact solid-state LiDAR with wider field-of-view — addressing miniaturisation barriers; Pix4D integrated Gaussian splatting into production desktop software, improving point cloud completeness in low-texture areas. LiDAR SLAM achieves ±1–2cm mapping accuracy with 76.76% drift reduction; multi-modal sensor fusion reaches 91.6% mAP with 80% weather retention in heavy fog and rain, signalling ecosystem transition from experimental to commercial AV deployment. Aerial imaging market confirmed at $4.2B (2025), projected $12.4B by 2034 (17.2% CAGR), with 68% of global aerial imagery now UAV-captured. Technology bifurcation hardened: photogrammetry scales profitably, LiDAR anchors autonomous mobility—but neither resolved weather robustness or cost accessibility barriers preventing mainstream adoption.
• 2026-Jun (mid-to-late): Vendor hardware maturation advanced with Ouster announcing Rev8 native color LiDAR (Fujifilm color science, 200m range at 10% reflectivity, 42.9 GMACs processing) with broad OEM adoption (Google, Volvo, Liebherr, Epiroc, Seegrid, Skydio, PlusAI, Gecko Robotics) and immediate shipping; concurrent Ouster BlueCity deployment confirmed 700+ contracted infrastructure sites and 40+ NJDOT highway locations for real-time traffic management. Artec Jet SLAM LiDAR GA with GPS-independent operation (±10mm accuracy indoor, ±15mm general, 1.9M points/sec) deployed across concert halls, mines, heritage sites, power lines, industrial facilities. Structural inflection confirmed: RoboSense Q1 2026 robotic LiDAR surged 1,458.8% YoY to 185,500 units (56% of total shipments), exceeding ADAS and spanning robotic lawnmowers, autonomous delivery, humanoid robots, cleaning systems across 50+ companies; RoboSense also secured a 500K-unit FAW Toyota production supply contract, confirming consumer-segment OEM adoption at scale. Production deployments demonstrate maturity: Hitachi excavator real-time digital twin achieved ±20mm accuracy, 45ms latency vs. 150–300ms competitors; Gecko Robotics deployed Rev8 for infrastructure inspection across Fortune 100 energy/O&G and U.S. military (Air Force, Navy, submarines); Niantic Spatial platform achieved DoW Tradewinds "Awardable" government procurement status with deployment across military services (ingesting drones, smartphones, ISR/UAS; 90% SPZ compression). PIX4D integrated 3D Gaussian Splatting across its product line with PIX4Dcatch achieving cm-level RTK accuracy from smartphones, further validating consumer-device viability for commercial surveying. Robustness research identified critical failure: solid-state LiDAR internal-multipath glare (phantom objects); ATLAS adversarial LiDAR benchmark revealed an additional robustness asymmetry—higher-benchmark models are more vulnerable to injection attacks, documenting architecture-agnostic deployment readiness gaps. Autonomous system evaluation documented severe limits: 98% success under isolated disturbances vs. 52% under combined conditions (fog+rain+night+traffic). Annotation emerged as a binding production constraint: Uber documented that enterprise multi-modal 3D labeling pipelines (geometry, calibration, occlusion handling) are now standard but remain the bottleneck for scaling autonomous systems. Vendor consolidation continued with Chinese suppliers dominant in automotive ADAS while global players (Ouster, Artec, Niantic) expanded into infrastructure, government, and industrial robotics. Late June 2026 scan context: Intel's Mobileye Drive announced 150k+ LiDAR unit commitment (9 sensors per vehicle, 17k vehicle scale target); Applied Intuition's camera-radar architecture achieved production deployment in multiple geographies without LiDAR, signaling competing architectural approaches. DJI Terra Gaussian Splatting integration demonstrated reconstruction algorithm maturity beyond classical photogrammetry. Aptiv and BitSensing vendor analyses documented specific LiDAR failure modes (warehouse reflective surfaces, adverse weather attenuation) and architectural necessity questions. Saudi Arabia's mega-project drone surveying adoption (NEOM, Qiddiya, Diriyah) confirmed photogrammetry as production-standard contract deliverable. These late-June signals reinforce the tier-defining tension: hardware/software/algorithmic maturity is advancing, yet sensor robustness, cost accessibility, and architectural necessity remain unresolved barriers constraining scaled mainstream adoption.