Score's Vision: Making Every Camera Intelligent

The world generates over 500 billion hours of video footage daily from 150+ billion cameras globally, yet 90% of it remains unlabeled and underused. While companies pour billions into AI language models, the visual world (factories, sports stadiums, petrol stations, hospitals) still operates largely blind to what their cameras actually see.

Score, a Vision AI company, is building what they call "Vision AutoML": a self-improving computer vision system that learns from competitive incentives rather than static datasets. Their mission is simple but radical: make every camera intelligent.

The numbers suggest they might be onto something. Score processes football games at $10 per match with 94% accuracy in just two minutes. Compare that to Microsoft Azure AI Vision charging $0.05 per minute of video plus $0.25 per 1,000 queries, or Google Cloud Vision API at $1.50 per 1,000 images. Enterprise solutions from Clarifai run $500+ monthly for millions of operations, while custom AI vision development starts at $30,000 and scales to $90,000+ for complex systems. At 100x to 1,000x cheaper than comparable solutions, Score isn't just competing on price. They're also unlocking entirely new markets.

Deloitte's 2025 tech value survey found that 74% of organizations invested in AI and generative AI over the past year, nearly 20 percentage points higher than any other technology capability. Digital budgets have surged from 7.5% of revenue in 2024 to 13.7% in 2025, with AI automation capturing roughly 36% of those budgets (about $700 million for a company with $13 billion in revenue). But there's a gap. While language models dominate AI investment, visual intelligence lags despite its massive potential across manufacturing, healthcare, automotive, retail, agriculture, and security.

Manufacturing sees ROI through improved quality control and reduced defects. Healthcare gains from enhanced diagnostics. Retail benefits from inventory optimization and automated checkout. Studies show AI vision solutions delivering over 600% ROI with payback in months while reducing manual labor by 75%. The bottleneck isn't compute or algorithms. It's accessible, production-ready systems that work on real-world footage without requiring ML expertise and massive labeled datasets.

Recently, Build AI's Eddy Xu released the largest egocentric dataset in history: 10,000 hours of first-person footage from 2,153 factory workers performing real manufacturing jobs. One billion frames of humans doing skilled manual labor, all open-sourced. That dataset represents the scale of visual data companies need to train embodied AI systems. Score is building the infrastructure to make that data actionable.

Traditional competitors process football games manually at $70 to $100 per match. Score does it for $10 in two minutes with 94% accuracy. "The only reason we're able to do this is because we've been optimizing for speed," explains Maxime Sebti, Co-founder and CEO of Score.

This matters at scale. Processing 100 football games costs $1,000 with Score versus $7,000+ with traditional services. But the real opportunity isn't sports. It's the 150 billion cameras globally generating footage that needs analysis. Petrol stations. Manufacturing facilities. Retail stores. Hospitals. Agriculture. Security systems. Every domain with cameras becomes addressable when the economics shift from $70 per analysis to $10.

Score began in sports (a domain defined by motion, precision, and stakes) but the same intelligence now scales to manufacturing, retail, healthcare, and urban systems. Recent partnerships demonstrate the spread.

Reading FC became the first professional football club to appoint a head of AI, partnering with Score on a one-year research collaboration to develop Vision AI agents. Stuart Fenton, Reading's head of AI, explains: "We see immense potential in co-developing these Vision AI Agents to revolutionize how we analyze performance and create the 'Reading Model.'" The system will analyze players across any league using objective assessments, deliver rapid match analysis, support tactical preparation, and expand into fan experience and retail operations. Brian McDermott, the club legend who led Reading to the Premier League and Championship title in 2012, is involved in the research collaboration, combining scientific expertise with footballing experience.

AVIA, one of Europe's leading petroleum distributors with 3,000+ petrol stations across 15 countries, is deploying Score's Vision AI to build "Smart Petrol Stations." As stations move toward full automation, Score's system detects operational issues in real time: customers unable to refuel due to equipment malfunction, broken payment terminals, leaking or jammed fuel pistols, blocked bays, irregular vehicle flow. "True autonomy requires intelligence," says Guillaume Boussaroque, AVIA's Network Manager. "Score allows us to add real-time awareness to our network, helping every station stay safe, operational, and responsive without human delay."

Score's stated mission is to democratize visual intelligence by building an open, permissionless computer vision layer. Their systems learn from competitive incentives through Bittensor's decentralized network, where they operate as "subnet 44" driving continuous improvement without relying on static datasets that become outdated. This approach inverts the traditional computer vision business model. Instead of selling expensive enterprise licenses that require dedicated ML teams, Score provides a platform where developers upload video and receive structured intelligence. The complexity (model selection, training, optimization, deployment) happens behind the scenes.

Score's partnerships with Reading FC and AVIA serve as proof-of-concept at opposite ends of the use case spectrum: high-performance sports analytics and autonomous infrastructure monitoring. If the same core technology works for both, the generalizability suggests Score has found something fundamental about how to process video intelligence at scale.

The world has 150 billion cameras generating 500 billion hours of footage daily. Most of it is recorded, stored, and ignored. Score is betting that shifting the economics from $1,000 per analysis to $10 changes who can afford to build with visual intelligence. From football pitches to petrol stations to factory floors, the infrastructure is finally here to make it accessible.