# Lot Vulture – Turn existing RTSP cameras into live parking occupancy maps

> Source: <https://github.com/lotvulture/lotvulture>
> Published: 2026-09-03 14:38:55+00:00

**Lot Vulture** turns any existing IP or RTSP camera into an automated, real-time parking lot intelligence system. *Circling the lot so you don't have to*, Lot Vulture performs on-device computer vision inference directly on your camera feeds—delivering live space-level occupancy maps, historical scrubbing timelines, customizable alerts, and deep parking utilization analytics without requiring expensive physical in-ground sensors.

| Challenge with Traditional Solutions | How Lot Vulture Solves It |
|---|---|
In-Ground Magnetic Sensors ($200–$500/spot): Expensive to core-drill, battery-dependent, and frequently destroyed by snowplows and asphalt repaving. |
Uses Existing Cameras: Turn standard CCTV, RTSP, or snapshot IP cameras into smart parking sensors with zero pavement alteration. |
Proprietary Sensor Hardware & Vendor Lock-In: Requires purchasing expensive proprietary single-vendor cameras or per-space overhead light indicators. |
Camera & Hardware Agnostic: Works with any standard RTSP, ONVIF, or snapshot IP camera from any manufacturer (Hikvision, Dahua, Axis, Hanwha, Ubiquiti, etc.). |
Noisy AI & False Flickers: Minor lighting shifts or tree shadows cause constant "space full / space free" notification spam. |
Dual-Threshold Hysteresis: Sticky state transitions ensure rock-solid, flicker-free dashboard data. |
Manual Lot Audits: Attendants driving golf carts with clipboards or radios providing only delayed, periodic estimates. |
Live Digital Dashboard & Historical Timeline: Real-time occupancy map, second-by-second scrubbing, and automated alerting. |

- 📍
**Interactive Visual Space Editor:** Define exact parking spots in seconds with a high-precision polygon canvas powered by Vue-Konva. Supports angled, parallel, and tight-row parking configurations. - ⚡
**High-Performance Edge AI Engine:** Powered by`vulturevision`

(C++ ONNX Runtime with native DirectML on Windows and CUDA on Linux) with automatic CPU fallback. - 🔭
**Visual Scrubbing History Timeline:** Scrub through every minute of the day with second-by-second precision. Click any space to see exact visual snapshots of state transitions. - ⚙️
**Smart Sensitivity (Hysteresis Logic):** Dual-threshold certainty eliminates status bouncing during dusk, rain, snow, or transient occlusions. - 🔔
**Multi-Channel Alerting Subsystem:** Trigger Webhooks, MQTT messages (ideal for Home Assistant & smart signage), or Emails on lot capacity thresholds, single space transitions, or camera health events. - 🔄
**Human-in-the-Loop "Improve AI" Feedback:** Operators can flag and correct misdetections directly from the UI to continuously feed the training pipeline. - 🛡️
**Role-Based Multi-User Access:** Granular roles for operators, managers, and view-only analytics dashboards.

Download the latest standalone installer (`LotVulture_Setup_*.exe`

) from the ** Releases** page.

- Packages the complete web dashboard, C++ DirectML runtime, and background Windows service daemon.
- No WSL, Docker, or external dependencies required.
- Launch the app or open
`http://localhost:8000`

in your browser.

Run the official container with GPU acceleration:

```
docker run -d \
  --name lotvulture \
  --restart unless-stopped \
  -p 8000:8000 \
  --gpus all \
  -v lotvulture-data:/var/lib/lotvulture \
  ghcr.io/lotvulture/lotvulture:latest
```

*(If running on a system without an NVIDIA GPU, omit --gpus all — Lot Vulture will automatically utilize optimized CPU inference.)*

Access the dashboard at ** http://localhost:8000**.

```
version: '3.8'

services:
  lotvulture:
    image: ghcr.io/lotvulture/lotvulture:latest
    container_name: lotvulture
    restart: unless-stopped
    ports:
      - "8000:8000"
    environment:
      - VULTURE_LV_HOME=/var/lib/lotvulture
    volumes:
      - lotvulture-data:/var/lib/lotvulture
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
```

Run with:

```
docker compose up -d
```

**Connect Your Cameras:** Add RTSP stream URLs or snapshot endpoints in the Cameras view.**Draw Parking Zones:** Use the visual canvas editor to map the four corners of each space.**Automate & Monitor:** Real-time occupancy states stream immediately over WebSockets to your live dashboard.**Trigger Actions:** Stream lot availability to digital LED road signs via MQTT, notify staff when capacity hits 95%, or log audit history.

Lot Vulture is source-available with a commercial tier:

| Feature | Community Edition | Commercial License |
|---|---|---|
Inference Model |
Bundled Standard Model | High-Accuracy Commercial Model |
Visual Space Editor & Live Dashboard |
✅ Included | ✅ Included |
Visual History Timeline & Scrubbing |
✅ Included | ✅ Included |
Webhooks & MQTT Alerts |
✅ Included | ✅ Included |
GPU Acceleration (CUDA / DirectML) |
✅ Included | ✅ Included |
Commercial Monitoring Services |
✅ Permitted (with baseline model) | ✅ Permitted (with licensed seats) |
Cloud Telemetry & Continuous Retraining |
— | ✅ Included |
Priority Support & Multi-Camera Scaling |
Community | Dedicated |

To activate a commercial license or start a free 14-day trial, navigate to the **License & Billing** tab in your dashboard.

Want on-demand parking space detection in your own application without managing on-prem infrastructure?

Use the ** Lot Vulture Developer Cloud API** ($0.02 / image):

```
curl -X POST "https://api.lotvulture.com/v1/detect" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "image=@parking_lot.jpg" \
  -F 'rois=[[[120,340],[250,340],[260,490],[110,490]]]'
```

Architecture deep dive, local environment setup, and C++ engine compilation.[Developer Guide](/lotvulture/lotvulture/blob/master/docs/DEVELOPMENT.md):Running unit, integration, and performance tests.[Testing Guide](/lotvulture/lotvulture/blob/master/docs/TESTING.md):Guidelines for community contributions and pull requests.[Contribution Guidelines](/lotvulture/lotvulture/blob/master/CONTRIBUTING.md):

Copyright (c) 2026 Lot Vulture, LLC. All rights reserved.

This software is licensed under the ** Lot Vulture Source License 1.0.0 (Adapted from PolyForm Shield License 1.0.0)**.

**Permitted Use:** Broad permission to run, modify, and utilize the software internally and for commercial monitoring services.**Noncompete Condition:** You may not use the software to build, distribute, or sell a competing product or managed service, nor distribute modified vision engines designed to bypass licensing.**Commercial Model:** High-accuracy commercial neural network weights and cloud attestation assets are licensed separately under the[Commercial Terms of Service](/lotvulture/lotvulture/blob/master/LICENSE.md#definitions).

**Lot Vulture** — *Making parking smarter, one spot at a time.*

[Website](https://lotvulture.com) • [Report an Issue](https://github.com/lotvulture/lotvulture/issues)