The Mission
Nothing But Net was born from the intersection of competitive basketball and computer vision engineering. The goal was to build a production-grade analytics platform that provides elite-level intelligence to any player through a modern web application and mobile companion.
1. Full-Stack Web Platform & Analytics Suite (Core Flagship)
The primary hub of Nothing But Net is a high-performance web application hosted at nothingbutnet.online. Built with modern full-stack standards, the web platform serves as the central control room for shot analysis, player telemetry, and platform management:
- Interactive React 19 Frontend: Engineered with React 19, Vite, and Tailwind CSS to deliver sub-second UI transitions, interactive shot charts, coordinate heatmaps, and customizable video scrubbing.
- Node.js API Gateway & Session Management: A secure Node.js/Express backend coordinating user authentication (JWT), MongoDB data persistence, and Stripe payment webhook processing for Free, Guest, and Pro tiers.
- Comprehensive Visual Telemetry: Renders 2D shot arcs, calculated entry angles relative to the hoop rim, release points, and historical shooting percentages across multi-session practice routines.
- Production CI/CD: Continuous deployment pipeline with automated preview and production builds on Vercel backed by resilient cloud infrastructure.
2. Native Android Mobile App & On-Court Companion
To bring the power of the web platform directly onto the basketball court, a dedicated native Android application was engineered in Kotlin as a specialized mobile recording companion:
- CameraX Recording & Frame Optimization: High-frame-rate on-court video capture configured with dynamic exposure and shutter speed tuning tailored for high-speed sports motion.
- Resilient Background Ingestion: Non-blocking background upload pipelines that buffer and stream recordings to the cloud microservices without locking the athlete out of reviewing prior sets.
- Sideline Shot Telemetry: Instant on-device feedback displaying make/miss classifications, release angle readouts, and synced session logs synchronized with the user's primary web profile.
3. Computer Vision Pipeline & Physics Filter
In outdoor basketball courts, lighting conditions change rapidly, and spectators or rim colors can easily fool classical color masks. To achieve >90% trajectory accuracy:
- Nearest-Circle Selection: Uses spatial proximity heuristics across sequential frames to maintain ball identity even through occlusions.
- Physics-Based Jump Filter: Discards any detection that jumps more than 300 pixels within a single frame interval—physically impossible for human shooting velocities.
- Polynomial Trajectory Modeling: Fits detected coordinates to a second-order polynomial curve, allowing precise calculation of the entry angle into the basket cylinder.
4. Microservices Architecture & Cloud Deployment
The entire ecosystem is deployed across a decoupled multi-service infrastructure: GitHub Actions triggers automated deployments to Vercel for the React frontend, while the containerized Node.js API and Python computer vision microservices run on Render with automated health check monitoring and non-blocking worker threads.