# Video Gait Intelligence Business Plan ## 1. Executive summary This project is a video-based animal gait analysis platform. The initial wedge is equine gait and lameness review support, because the current technical stack, data, and validation workflow are horse-first. The larger market expansion is livestock monitoring, especially dairy cattle, where lameness is a high-frequency welfare and productivity problem. The product should not be positioned as an autonomous veterinary diagnosis system. The safer and more commercially realistic position is: > Objective video gait measurement, screening, and longitudinal trend monitoring for veterinarians, trainers, farms, and animal health operators. The core asset is a 3D skeleton sequence engine trained from synthetic anatomy, multi-camera real video, and eventually unlabeled fixed-camera videos. The long-term defensibility comes from animal-specific skeleton models, temporal biomechanical priors, pseudo-label pipelines, and real-world gait datasets. ## 2. Problem Lameness and gait abnormalities are expensive, under-detected, and inconsistently measured across horses, cattle, and pigs. ### Horses Equine lameness evaluation is clinically important but often subjective, expensive, and dependent on expert availability. High-value sport and racing horses justify premium review tools, but the market size is narrower than cattle. Primary pain points: - Subjective visual assessment varies by observer. - Repeated recovery monitoring is hard to quantify. - Owners and trainers want earlier warning before performance loss. - Clinics need clearer visual reports for communication. ### Dairy cattle Cattle lameness is a major economic and welfare issue. Reviews commonly describe dairy cow lameness as prevalent, costly, and under-detected, with reported herd prevalence often around 10-30% and median prevalence around 22% in reviewed studies. Primary pain points: - Large herd size makes manual observation unreliable. - Lame cows lose productivity and fertility and have higher culling risk. - Farmers need early alerts, not late-stage diagnosis. - Welfare and audit pressure increases demand for objective monitoring. ### Pigs Swine lameness is also commercially relevant, especially in sows and grower-finisher pigs. The technical environment is harder because of group housing, occlusion, and individual identity tracking. Primary pain points: - Lameness affects welfare, growth, reproduction, and culling. - Group pens make individual gait measurement difficult. - Automated monitoring is more likely to be sold as health/welfare anomaly detection than as high-precision 3D diagnosis. ## 3. Solution The platform converts ordinary or fixed camera animal video into temporally stable 3D skeleton sequences and gait metrics. Pipeline: ```text Video -> animal detection / crop tracking / optional segmentation -> 2D and visual feature extraction -> 3D skeleton sequence model -> temporal correction + anatomical constraints -> gait cycle metrics and asymmetry signals -> report / alert / trend dashboard ``` Initial horse output: - 3D skeleton overlay - stride and gait cycle segmentation - head/pelvis vertical asymmetry candidates - distal limb trajectory and timing - left/right asymmetry trend - confidence and failure-mode flags - veterinarian-facing report Livestock output: - per-animal lameness risk score - mobility trend - abnormal gait or activity alert - herd-level welfare dashboard - review queue for farm staff or veterinarian ## 4. Product strategy ### Phase 1: Equine vet review report Objective: earn early revenue and prove technical value in a premium niche. Customer: - equine veterinarians - racing stables - sport horse trainers - rehabilitation centers Product: - upload video - receive gait skeleton overlay and objective metrics - longitudinal comparison for the same horse - exportable vet report Commercial position: - "Decision support and measurement" - not "automatic diagnosis" Likely pricing: - per report: USD 20-100 - clinic plan: USD 300-1,500/month - enterprise/racing stable: custom ### Phase 2: Dairy cattle lameness monitoring Objective: enter the larger market. Customer: - dairy farms - herd management platforms - animal health companies - farm camera / sensor vendors Product: - fixed-camera monitoring lane - mobility score and lameness risk - early alert and review queue - integration with farm management software Likely pricing: - per farm monthly SaaS - per camera / per herd pricing - API licensing to livestock monitoring vendors ### Phase 3: Swine welfare and anomaly monitoring Objective: expand from gait analysis into group animal welfare monitoring. Customer: - large swine producers - welfare audit providers - farm automation vendors Product: - top-down or side-view pen monitoring - abnormal movement / posture / activity alerts - sow lameness risk where individual tracking is possible Positioning: - broader welfare and health monitoring, not only 3D lameness. ## 5. Market assessment Available market reports suggest animal health and livestock monitoring are large and growing categories: - Global livestock monitoring market: estimated USD 5.73B in 2025, projected USD 14.82B by 2033. - Bovine monitoring market: estimated USD 2.7B in 2025, projected USD 7.1B by 2033. - Equine healthcare market: estimated USD 4.56B in 2025, projected USD 10.40B by 2033. - Veterinary diagnostics market: estimated USD 8.4B in 2025, projected USD 19.2B by 2033. - Veterinary AI diagnostics market: estimated USD 652M in 2024, projected USD 2.63B by 2033. These are broad markets. The directly serviceable market for video gait analysis is much smaller. The most credible wedge is a paid equine gait report product, followed by cattle lameness monitoring integrations. ## 6. Competitive landscape Competitors and substitutes: - manual veterinary visual assessment - IMU-based lameness systems - pressure mats / force plates - farm activity collars and pedometers - camera-based livestock monitoring companies - general computer vision vendors Differentiation: - video-only or camera-first workflow - 3D skeleton sequence, not just 2D keypoints - temporal biomechanics and anatomy constraints - synthetic-to-real training pipeline - ability to learn from unlabeled fixed-camera video - animal-specific expansion path Weaknesses versus alternatives: - lower precision than force plates or IMUs for controlled clinical measurement - real-world video is sensitive to occlusion, camera angle, crop quality, and lighting - regulatory and clinical claims must be conservative ## 7. Technical roadmap ### V24: temporal horse skeleton Goal: - reduce temporal jitter and distal limb instability. Proof: - synthetic sequential validation - velocity, acceleration, bone consistency, MPJPE ### V25: camera and dataset diversity Goal: - reduce dependence on exact fixed camera setup. Proof: - hold accuracy across wider camera/crop distributions - avoid regression when scaling pose/shape/camera diversity ### V25.5: LeWorldModel-lite Goal: - learn a dynamics prior for plausible animal skeleton motion. Use: - temporal critic - invalid transition scoring - pseudo-label filtering ### V26: 10-camera, 5-horse real-video bridge Goal: - measure sim-to-real gap and create pseudo-3D seed labels. Use: - multi-view consistency - camera calibration and sync checks - real-video failure catalog ### V27: GT-free video learning Goal: - train with unlabeled fixed videos using weak/self-supervised signals. Signals: - 2D reprojection - segmentation/silhouette consistency - bone constancy - temporal consistency - LeWorldModel score - pseudo-label confidence ### V28: YouTube-heavy fixed-video robustness Goal: - stable skeletons on ordinary fixed horse videos. Constraint: - synthetic 3D ground truth remains as a regression anchor. - YouTube-only training is not recommended. ## 8. Data strategy Current horse data assets: - synthetic hSMAL/PFERD training data - true sequential synthetic cache for V24 - planned 10-camera, 5-horse real dataset Near-term data needs: - real horse video ingestion and tracking labels - camera calibration/sync metadata - veterinarian review labels for lameness severity or confidence - failure-case library Expansion data needs: - cattle skeleton definition and real farm video - cattle lameness labels or locomotion scores - pig pen videos with individual tracking or group anomaly labels Key principle: - 3D ground truth is required early. - Later unlabeled video can dominate training, but synthetic or curated GT anchors should remain to prevent drift. ## 9. Business model ### Equine Revenue: - pay-per-report - clinic subscription - stable/team subscription - API for equine platforms Buyer: - veterinarians - trainers - horse owners - racing/sport organizations Sales motion: - founder-led sales - pilot with clinics and stables - publish before/after recovery case studies ### Cattle Revenue: - SaaS per farm - per camera / per herd pricing - integration licensing - analytics API Buyer: - farm operators - dairy groups - herd management software vendors - animal health companies Sales motion: - partnerships are likely more efficient than direct farm-by-farm sales. ### Swine Revenue: - welfare monitoring module - enterprise farm contracts - integration with barn camera systems Buyer: - large producers - welfare/audit providers - farm automation vendors ## 10. Go-to-market plan ### 0-3 months Milestones: - complete V24 full temporal validation - generate strong horse skeleton overlay demos - produce 5-10 report examples - define vet-facing metrics and report format Commercial goal: - secure 3-5 expert reviewers or pilot clinics. ### 3-6 months Milestones: - run V26 real-video bridge on 10-camera, 5-horse dataset - compare synthetic vs real failure modes - create first paid or design-partner equine report workflow Commercial goal: - paid pilots or formal letters of intent. ### 6-12 months Milestones: - V27 GT-free learning prototype - real fixed-video robustness demos - start cattle proof-of-concept with partner data Commercial goal: - equine subscription pilots - cattle partner pilot ### 12-24 months Milestones: - cattle lameness monitoring MVP - farm integration - larger unlabeled video training pool Commercial goal: - recurring B2B revenue through equine clinics and cattle pilots. ## 11. Operating model These are planning assumptions, not guaranteed forecasts. They are included to make the investment case testable. ### Pricing assumptions | Product line | Early price assumption | Notes | |---|---:|---| | Equine single video report | USD 20-100 / report | Useful for early paid validation and expert review workflows | | Equine clinic subscription | USD 300-1,500 / month | Depends on report volume, history, and collaboration features | | Stable / racing team plan | custom | Higher willingness to pay if longitudinal monitoring is tied to performance and recovery | | Cattle monitoring SaaS | USD 6,000-30,000 / farm / year | Depends on herd size, camera count, and integration depth | | API / platform licensing | custom | Best path if sold through herd management or camera vendors | ### Three-year revenue scenario | Year | Revenue scenario | Main assumption | |---|---:|---| | Year 1 | USD 0.1-0.3M | 20-40 equine clinics/stables, report workflow validation, limited subscription conversion | | Year 2 | USD 0.7-1.5M | 100-250 equine paid accounts plus first cattle paid pilots or integration revenue | | Year 3 | USD 3-7M | Cattle monitoring becomes the main growth driver; equine remains premium proof and data channel | ### Milestone gates | Gate | Must prove | Decision | |---|---|---| | V24 gate | Stable temporal skeletons on synthetic sequence data | Continue equine report demo | | V26 gate | 10-camera real-video bridge produces reliable overlays and pseudo-3D seed labels | Start paid equine pilots | | V27 gate | Unlabeled fixed-video learning improves robustness without drift | Expand video pool and start cattle PoC | | Cattle PoC gate | Farm videos produce useful mobility risk ranking | Pursue cattle SaaS / platform partnerships | ### Failure threshold If V26 cannot produce reliable real-video horse reports under controlled fixed-camera conditions, the company should pause cattle expansion and focus on data quality, camera calibration, and the real-video ingestion pipeline. The business should not scale before the real-video bridge is credible. ## 12. Key risks ### Technical risks - Synthetic-to-real gap is larger than expected. - Real video crop/tracking quality dominates skeleton quality. - 5 horses are insufficient for real-world generalization. - YouTube pseudo-labels collapse without strong anchors. - Moving cameras and occlusion remain hard. Mitigation: - keep synthetic GT anchor - use 10-camera real seed for pseudo-3D - report confidence and failure states - start with fixed-camera use cases ### Clinical and regulatory risks - Automated diagnosis claims may create liability. - Vet adoption requires trust and explainability. - Clinical validation is expensive. Mitigation: - position as decision support and trend monitoring - keep veterinarian in the loop - avoid medical diagnosis claims until validated ### Commercial risks - Equine market may be premium but narrow. - Farm sales cycles can be slow. - Hardware installation can reduce adoption. Mitigation: - start video-upload first for equine - partner for cattle distribution - use existing camera infrastructure where possible ## 13. Objective recommendation The best business path is: ```text Horse first for technical proof and premium early revenue. Cattle second for the largest scalable market. Pig later as a welfare/anomaly monitoring expansion. ``` The company should not present itself as a generic animal AI company at the start. It should present a focused product: > Video-based objective gait measurement for equine veterinarians and performance horse teams. Then expand the same skeleton, temporal, and weak-label learning stack into: > Camera-based cattle mobility and lameness monitoring. This balances feasibility, credibility, and market size. ## 14. Evidence and sources - [Grand View Research, Livestock Monitoring Market](https://www.grandviewresearch.com/industry-analysis/livestock-monitoring-market): USD 5.73B in 2025, USD 14.82B projected by 2033. - [Grand View Research, Bovine Monitoring Market](https://www.grandviewresearch.com/industry-analysis/bovine-monitoring-market-report): USD 2.7B in 2025, USD 7.1B projected by 2033. - [Grand View Research, Equine Healthcare Market](https://www.grandviewresearch.com/industry-analysis/equine-healthcare-market-report): USD 4.56B in 2025, USD 10.40B projected by 2033. - [Grand View Research, Veterinary Diagnostics Market](https://www.grandviewresearch.com/industry-analysis/veterinary-diagnostics-market): USD 8.4B in 2025, USD 19.2B projected by 2033. - [Grand View Research, Veterinary AI Diagnostics Market](https://www.grandviewresearch.com/industry-analysis/veterinary-ai-diagnostics-market-report): USD 652M in 2024, USD 2.63B projected by 2033. - [Animals 2021 cattle lameness review](https://www.mdpi.com/2076-2615/11/11/3033): cattle lameness is a major welfare and productivity issue in precision livestock farming. - [Veterinary Journal 2023 dairy cow lameness prevalence review](https://www.sciencedirect.com/science/article/pii/S1090023323000266): median lameness prevalence reported around 22% across reviewed studies. - [BMC Veterinary Research 2014 pig lameness gait analysis](https://bmcvetres.biomedcentral.com/articles/10.1186/s12917-014-0193-8): pig lameness affects welfare and farm economics and can be objectively quantified through gait asymmetry. - [Merck Veterinary Manual swine lameness overview](https://www.merckvetmanual.com/musculoskeletal-system/lameness-in-pigs/overview-of-lameness-in-pigs): swine lameness affects growth, reproduction, viability, and herd-level production.