The Vision Suite

The Computer Vision Flywheel

From raw video capture to high-frequency hardware execution — annotate datasets, train custom models, test your rig, and deploy with sub-millisecond precision.

1. Label & Curate YOLO11 · SAM 2

SentinelLabeler

AI-assisted dataset labeling built specifically for live game-capture cards. Features dual-model live compare, temporal box propagation across whole clips, and SAHI tiled inference.

Temporal propagation (label 1 frame, track clip)
Live dual-model A/B comparison on real footage
Exports clean YOLOv8/YOLO11 format datasets

Launch SentinelLabeler →

2. Model Hub DirectML · TensorRT

Model Hub & Catalog

Curated and creator-submitted detection architectures tuned for high refresh rates. Hot-swap models on the fly directly inside the launcher without restarting your session.

Pre-compiled ONNX & TensorRT engines
Creator storefront: upload & monetize models
Zero-downtime hot swapping in active engine

Browse Models →

3. Real-Time Engine 240Hz · 1,000Hz Loop

Sentinel Core Vision

Zero-copy DXGI desktop ingest, real-time GPU inference, and multi-tier visual servoing kinematics delivering precision outputs to ViGEm virtual controllers or hardware bridges.

Zero-copy GPU VRAM shared texture ingest
1,000Hz multimedia motor loop (sub-0.3ms jitter)
Live localhost tuning studio & diagnostic HUD

Explore Engine Features →

Silicon Benchmarks

Engineered for Silicon. Tested on Real Metal.

Empirical inference latency and frame delivery rates across modern GPU architectures on 320x320 and 640x640 vision networks.

HARDWARE PERFORMANCE BENCHMARK MATRIX

Test Your Rig With Free Profiler →

GPU Silicon Architecture	Execution Provider	Inference Latency	Throughput (FPS)	Capture Speed	Competitive Grade
■ NVIDIA RTX 4090 / 4080	TensorRT (FP16)	1.4 ms	714 FPS	240+ FPS (DXGI)	Tier 1 · Ultra Pro
■ NVIDIA RTX 5050 Laptop (Tested)	TensorRT / DirectML	2.9 ms	345 FPS	240 FPS (DXGI)	Tier 1 · Esports Ready
■ NVIDIA RTX 3070 / 3060	DirectML (DX12)	4.8 ms	208 FPS	165 FPS (DXGI)	Tier 1 · High Refresh
■ AMD Radeon RX 7800 XT / 6700	DirectML (DX12)	5.4 ms	185 FPS	144 FPS (DXGI)	Tier 2 · Pro Grade
■ Intel Arc A770 / Iris Xe	OpenVINO / DirectML	7.6 ms	131 FPS	120 FPS (DXGI)	Tier 2 · 120Hz Smooth

The pipeline, in motion

Watch every frame become an action

Inside Sentinel, data moves as discrete packets — captured, understood, fused, tracked, chosen, and driven to output. Each stage runs at its own speed and reads only the freshest truth, so nothing waits on anything it doesn't need.

Capture

Frames

Preprocess

Normalize

Fusion

Model + color

Tracker

Identity

Target Selector

Priority

Motion Engine

Trajectory

Motion Output

Device

Live packets flowing between stages — capture to output in milliseconds

The pipeline

Every stage, in one engine

From frame to action in milliseconds — and every layer is yours to inspect and tune.

Detection

Fast ONNX inference scans every frame with GPU acceleration.

Multi execution provider · CUDA / TensorRT / DirectML / OpenVINO

Tracking

Persistent object identities with predictive motion.

Multiple swappable engines

Vision view

Watch every detection and decision drawn live as it happens.

Feed-first overlay

Color logic

Optional HSV fusion confirms the model in real time.

Live color scout

Models

Hot-swap detection models on the fly — no restart.

Drop-in model manager

Output

Low-latency output to a supported device, delivered as native input.

Native input device

Why it exists

Stop guessing at a black box

Most vision setups hide what the machine is doing. Sentinel makes every layer visible and tunable, live.

Traditional setups

✕ Guesswork and hidden settings
✕ Restart after every change
✕ No visual feedback on decisions
✕ Config files and command lines

Sentinel Core Vision

✓ Every layer visible, drawn live
✓ Instant tuning in your browser
✓ Local, on-device processing
✓ One dashboard, zero config files

The workflow

Frame to tuned in five steps

1

Install

Run the launcher — engine and dashboard unpack in seconds.

2

Activate

Paste your key. It binds to your PC and activates instantly.

3

Choose a model

Pick a detection model — hot-swap any time.

4

Tune the pipeline

Dial in the feel from simple sliders to full developer surfaces.

5

Run

Watch it work live and adjust on the fly.

Inside the platform

Experience the platform live

Click between real-time vision servoing, dataset labeling with SentinelLabeler, empirical silicon benchmarks, and zero-GC developer APIs.

INGEST: 240 FPS (DXGI) INFERENCE: 2.9 ms (FP16) TIMER JITTER: 0.18 ms ACTIVE TARGET: #104 (ADS LOCKED)

Ingest Rate

240 FPS

GPU Latency

2.9 ms

Hardware Bus

1,000 Hz

Timer Slew

0.18 ms

Kinematic Visual Servoing IMM FILTER ON

Time-optimal phase-plane snapping automatically leads moving targets without overshoot or jitter. Click any target on the canvas to switch lock.

FOV Search Radius150 px

Servo Snap Stiffness12%

Interactive real-time preview • Zero CPU allocations

SAM 2 • 1-CLICK POLYGON SEGMENTATION

Click anywhere on the object to reposition SAM 2 prompt point

Dataset Annotation Engine

From Raw Capture to YOLO in Minutes

Built specifically for capture card video streams. SentinelLabeler eliminates tedious manual frame-by-frame labeling with cutting-edge AI assistance:

⚡

Segment Anything 2 (SAM 2)

Click once on any target to instantly generate tight polygon segmentation masks with high edge fidelity.

⏩

Temporal Frame Propagation

Annotate a single frame and propagate bounding boxes across up to 300 sequential video frames automatically.

📦

Direct YOLOv8 & YOLOv11 Export

Outputs clean normalized YOLO text labels and train/val directories ready for direct PyTorch / ONNX training.

Open SentinelLabeler

2.9 ms

Mean Neural Inference Latency

Tier 1 · Esports Grade (144Hz–240Hz+ Ready)

Architecture NVIDIA GeForce RTX 5050 Laptop GPU

Execution Provider TensorRT / DirectML (FP16 Engine)

Peak Throughput 345 FPS

Video Capture 240 FPS (DXGI Desktop Duplication)

Timer Jitter 0.32 ms (Multimedia 1000Hz)

Download Free Benchmark Tool → 105MB Portable • 0 Dependencies

// ============================================================================
// Sentinel Core Vision — Zero-GC 1,000Hz Hardware Dispatch
// High-Frequency Visual Servoing & Kinematic State Estimation
// ============================================================================

[StructLayout(LayoutKind.Sequential, Pack = 1)]
public readonly record struct DirectHidPacket(
    byte ReportId,       // 0x01 = Direct Controller Emulation
    short StickLeftX,    // [-32768, 32767] Normalized Deflection
    short StickLeftY,
    short StickRightX,   // Servoing Saccade Step
    short StickRightY,
    uint ButtonsMask,    // Raw Gamepad Triggers & ADS
    uint TimestampMicro  // 1,000Hz Jitter Compensation
);

public sealed class ServoKinematicPipeline : IDisposable
{
    private readonly OutputLoop _outputLoop;
    private readonly ProjectionMagnetCleanAimCore _aimCore;

    [MethodImpl(MethodImplOptions.AggressiveInlining | MethodImplOptions.AggressiveOptimization)]
    public void DispatchServoTick(in ReadOnlySpan targets, in VideoFrameHeader frame)
    {
        // Zero allocation pass: IMM filter estimates target velocity & acceleration
        ref readonly var activeTarget = ref _aimCore.EvaluateBestTarget(targets, frame.Timestamp);
        
        // Time-optimal phase-plane snapping with anti-deadzone compensation
        Vector2 stickDeflection = _aimCore.ComputeKinematicStep(activeTarget);
        
        // Zero-GC atomic transmission to Titan Two hardware via USB DirectHID
        _outputLoop.EnqueueStickDeflection(stickDeflection.X, stickDeflection.Y);
    }
}

View all 6 native studio window views ▾

Dashboard

Live dashboard

System readiness and pipeline state at a glance.

Vision · live truth

Vision overlay

Raw feed and AI overlay, drawn live as it decides.

Color detection

Color logic

A live color scout plus model-and-color fusion.

Models

Model manager

Inspect and hot-swap detection models.

Setup

Devices & setup

Capture source, output device, calibration — checked for you.

Developer

Diagnostics

Runtime health, performance, and pipeline state.

Marketplace

A growing catalog of models, from us and from creators

Browse detection models built by Sentinel and by independent creators, hot-swap them straight into the engine, and check out a creator's own storefront before you buy.

Browse the catalog

Search and filter models by task, size, and creator — every listing hot-swaps straight into the engine.

Open the marketplace →

Creator storefronts

Every creator gets their own customizable storefront — grid, card, or compact — to showcase what they've published.

Browse creators →

Become a creator

Apply from your account in minutes. Once approved, publish models and stand up your own storefront.

Apply to become a creator →

Spotlight

Featured this month

Ranked by real download counts from the marketplace — not picked, earned.

Pricing

Simple plans. Full power.

Start monthly, or own it for life. Both unlock the full platform and every update.

Monthly

Full access, billed month to month. Cancel whenever.

$20/ month

Everything unlocked · updates included

Full detection + tracking engine
Multiple engines & color fusion
Local tuning dashboard
Discord support

Start monthly

Save 33%

3-Month

Best monthly rate — billed once every three months.

$40/ 3 mo

≈ $13.33 / month · billed quarterly

Everything in Monthly
Best effective monthly price
Billed every 3 months
Cancel anytime

Get 3-Month

Best value

Lifetime

Pay once, own it forever — including every future update.

$250one-time

No subscription · all future updates

Everything in Monthly
One-time payment, no renewals
All future engines & features
Priority Discord support

Get lifetime

Secure Stripe checkout Cancel anytime Runs 100% on your machine Instant key delivery

Looking for the full breakdown? Compare plans →

FAQ

Answers, before you ask

What exactly is Sentinel Core Vision?

A real-time computer-vision platform with a local tuning dashboard. It detects and tracks targets from your video feed, draws every decision live, and sends low-latency output through your hardware — all configurable in the browser.

What do I need to run it?

A Windows PC. Sentinel supports multiple execution providers — NVIDIA (CUDA / TensorRT), AMD & Intel (DirectML), and OpenVINO — and auto-selects the fastest one available; a CPU fallback exists but isn't recommended for real-time use. You'll also want a capture source and a supported output device. Full requirements are on the Downloads page.

Is my data sent anywhere?

No. Every frame is processed entirely on your machine. The only network call is a lightweight check to the license server to confirm your subscription.

Monthly or lifetime — what's the difference?

Both unlock the complete platform and all updates. Monthly bills each month and cancels anytime; Lifetime is a single payment with no renewals.

Read all FAQs →

See every frame.Analyze in real time. Dominate the loop.

Real motion. Real-time overlay.

Watch the Sentinel Engine smooth the noise

The Computer Vision Flywheel

SentinelLabeler

Model Hub & Catalog

Sentinel Core Vision

Engineered for Silicon. Tested on Real Metal.

Watch every frame become an action

Every stage, in one engine

Detection

Tracking

Vision view

Color logic

Models

Output

Stop guessing at a black box

Traditional setups

Sentinel Core Vision

Frame to tuned in five steps

Install

Activate

Choose a model

Tune the pipeline

Run

Experience the platform live

From Raw Capture to YOLO in Minutes

Segment Anything 2 (SAM 2)

Temporal Frame Propagation

Direct YOLOv8 & YOLOv11 Export

Live dashboard

Vision overlay

Color logic

Model manager

Devices & setup

Diagnostics

A growing catalog of models, from us and from creators

Browse the catalog

Creator storefronts

Become a creator

Featured this month

Simple plans. Full power.

Answers, before you ask

Ready to see every frame?

See every frame.
Analyze in real time. Dominate the loop.