Build real-time vision systems with a performance-first C++ framework.
VectorSight is an upcoming C++ framework focused on low-latency computer vision pipelines and intelligent input assistance—designed for performance-critical applications.
- Focus
- Low latency
- Language
- Modern C++
- Targets
- Cross-platform
01 :: PIPELINE
Composable stages, predictable latency, and tight control over memory.
02 :: INPUT
Helpers that integrate vision signals with input workflows.
03 :: TOOLING
Examples, benchmarks, and docs tailored for real-world constraints.
Expressive & Type-Safe
Build pipelines that read like a story. VectorSight leverages modern C++ concepts to ensure type safety without sacrificing performance.
#include <vectorsight/pipeline.hpp>
int main() {
// Create a high-performance pipeline
auto pipeline = vs::Pipeline::create()
.source(vs::Camera::default_input())
.process(vs::Resize({1920, 1080}))
.process(vs::DetectObjects("yolo_v8"))
.sink(vs::Window::create("Preview"));
// Run with zero-copy where possible
pipeline.run();
}Designed for real-time constraints
The goal is to keep control in your hands: performance profiles, explicit tradeoffs, and a clean C++ developer experience.
Pipeline-first architecture
Compose stages, control boundaries, and keep latency measurable.
Performance primitives
A focus on explicit memory, batching, and predictable costs.
Input assistance layer
An opinionated layer for integrating vision signals with input workflows.
Cross-platform by default
Target multiple OS environments while keeping behavior consistent.
Benchmarks & examples
Realistic samples aimed at shipping quality—plus performance notes.
Clean, focused API
Avoid magic. Prefer clear defaults and explicit knobs when needed.
Roadmap (high level)
This is a living plan. The fastest way to influence it is by joining the community and sharing what you’re building.
-
Phase 1
Core pipeline + benchmarks
Establish core abstractions and baseline performance targets.
-
Phase 2
Input assistance utilities
Explore helpers that connect vision outputs to input workflows.
-
Phase 3
Docs, examples, stabilization
Improve DX: clear docs, recipes, and hardened APIs.
FAQ
Quick answers to common questions. If you have a specific use case, Discord is the best place to chat.
Is VectorSight open source?
The GitHub repo exists today, and the direction is to share work as it’s ready. Join Discord for the latest.
What makes it different?
The focus is latency + control: clear abstractions, performance-aware defaults, and intentional constraints.
When is the launch?
No hard date yet. Updates will be posted on Discord and GitHub.
How can I help?
Share your constraints: target FPS, resolution, hardware, OS, and pipeline details. That feedback shapes priorities.
Join the early community
Get announcements, share use cases, and help shape the roadmap.