Nemo Documentation

Getting Started

Welcome to Nemo, the AI-powered maritime hazard detection system. This guide will walk you through all features and settings to help you get the most out of your safety system.

System Requirements

• Operating System: Windows 10 or later
• Video Input: Any USB webcam, IP camera (RTSP), streaming feed, local video source, or onboard camera system
• Storage: ~2-4 GB free disk space
• Network: Optional — for multi-device access and Home Assistant integration. Nemo runs fully offline

Versions

Nemo is available in two versions:

• Intel version: Runs on any modern laptop or desktop. If you have CUDA already installed on your system, this version will likely use it automatically.
• CUDA version: Optimized for systems with a dedicated NVIDIA GPU and CUDA toolkit installed.

If you're unsure which to pick, start with the Intel version.

Quick Start

1. Run the Nemo installer (.exe)
2. Click through the license agreement and readme
3. Nemo installs to C:\Program Files\Nemo and creates two shortcuts on your desktop
4. Open the Nemo Server shortcut first
5. Then open the Nemo Dashboard shortcut to access the interface

Installation

Nemo ships as a standalone installer — everything is included. No Python, no pip, no Docker. Install and run.

Step 1: Download the Right Version

After purchasing your license, you'll receive a download link. Choose the version for your system:

If you're unsure, start with the Intel version.

Step 2: Run the Installer

1. Double-click the Nemo .exe installer
2. Click Next to accept the license agreement
3. Click Next to view the readme
4. The installer will install Nemo to C:\Program Files\Nemo
5. Two shortcuts are created on your desktop: Nemo Server and Nemo Dashboard

Step 3: Connect Camera

Plug in your USB camera before running Nemo.

Step 4: Start the Server

Open the Nemo Server shortcut on your desktop. The server must be running before you can use the dashboard.

Step 5: Open the Dashboard

Open the Nemo Dashboard shortcut on your desktop. This opens the Nemo interface in your web browser.

Step 6: Accessing from Another Device

To access the dashboard from a phone, tablet, or another computer on the same network:

1. Look at the Nemo Server window — it displays your IP addresses
2. You'll see 127.0.0.1 (this is localhost, only for the computer running Nemo)
3. Below that, you'll see another IP address (e.g. 192.168.1.100) — this is the one to use from other devices
4. On the other device, open a browser and go to http://[that-ip-address]:5001

No WiFi Network on the Boat?

Most boats don't have a WiFi network, which means your phone and other devices can't connect to the laptop running Nemo. To fix this:

1. On your laptop, turn on the Mobile Hotspot (Windows Settings > Network & Internet > Mobile Hotspot)
2. Connect your phone, tablet, or other devices to the laptop's hotspot
3. Then start the Nemo Server
4. Use the IP address shown in the server window to access the dashboard from your connected devices

Performance Tip

You'll get better performance out of Nemo if you don't have the dashboard open on the same machine. Once you've configured your settings, close the dashboard and let the server run in the background — it will continue detecting and alerting without the dashboard.

If you want to tune the system before heading out, you can point Nemo at a local video file path instead of a live camera. This lets you run recorded footage through the detection system to dial in your settings before you're on the water.

Verify Detection

You should see the live camera feed with detection overlays. Check the FPS counter in the header — it should read 15+ FPS for normal operation.

Updating Nemo

Updates are included for 1 year with your license. When a new version is available:

1. Download the new installer from your account
2. Close the running Nemo Server
3. Run the new installer — it will update the existing installation

Compatibility Check Tool

Before purchasing a software license, use the free Compatibility Check tool to verify your hardware can run Nemo.

What It Does

• Downloads and runs a stock RT-DETR model (public weights — not the custom Nemo maritime model)
• Performs real-time person detection on your USB camera feed
• Measures inference FPS on your specific hardware
• Gives you a clear pass or fail result

How to Run

1. Download the compatibility check tool from mynemo.ca/compatibility-check
2. Extract the archive for your platform
3. Connect a USB camera to your Jetson or laptop
4. Run the tool:

   Linux / Jetson:

   ./nemo-compat-check

   Windows:

   nemo-compat-check.exe

5. Watch the live FPS counter and detection output

Interpreting Results

Troubleshooting Failed Results

• Low FPS: Your GPU may not have enough compute power. Check that you're using a Jetson Orin or a laptop with a dedicated NVIDIA GPU.
• CUDA errors: Ensure NVIDIA drivers and CUDA toolkit are properly installed. On Jetson, use JetPack SDK.
• Camera not found: Verify the camera is connected and recognized by your OS (ls /dev/video* on Linux).
• Model not loading: Ensure the full archive was extracted — the model weights are bundled in the package.

Interface Overview

The Nemo interface is divided into several key areas:

Header Bar

Main Display

The central area shows the live camera feed with detection overlays. Key information displayed:

• Detection boxes: Colored rectangles around detected objects
• Risk indicators: Red highlighting for collision risks
• Tracking IDs: Unique identifiers for each tracked object
• On-screen stats: FPS, object count, and risk count

Warning Banner

When collision risks are detected, an orange/red warning banner appears at the bottom of the screen showing the number of potential risks.

Live Camera Feed

The live video feed is the heart of Nemo's detection system. Nemo accepts any video source — USB cameras, IP cameras, RTSP streams, streaming feeds, or local files.

Feed Controls

RAW vs Processed View

Processed View (Default): Shows the camera feed with image enhancements applied (gamma correction, CLAHE, contrast adjustments). This is what the AI sees and is optimized for detection accuracy.

RAW View: Shows the unmodified camera output. Useful for troubleshooting camera issues or comparing detection performance.

Detection Overlays

Objects are highlighted with colored bounding boxes:

• Green: Tracked object, no immediate risk
• Yellow: Object approaching, monitor closely
• Red: Collision risk detected

Detection System

Nemo uses advanced AI models trained specifically for maritime environments to detect and classify objects in real-time.

Detected Object Types

How Detection Works

1. Image Capture: Camera frames are captured at up to 60 FPS
2. Preprocessing: Images are enhanced for optimal detection (gamma correction, contrast adjustment)
3. AI Inference: TensorRT-optimized neural network analyzes each frame
4. Object Classification: Detected objects are classified by type
5. Risk Assessment: Object trajectories are analyzed for collision risk
6. Alert Generation: Visual and audio alerts are triggered when risks are detected

Collision Warnings

Nemo continuously analyzes object trajectories and positions to identify potential collision risks.

Warning Levels

Warning Banner

When one or more collision risks are detected, a warning banner appears at the bottom of the screen:

The banner shows the total number of active collision risks. Click the banner to highlight the specific objects causing warnings.

Risk Calculation Factors

Nemo considers multiple factors when assessing collision risk:

• Distance: Proximity of detected object to your vessel
• Relative velocity: Speed at which the gap is closing
• Bearing change: Whether the object's relative position is changing
• Object size: Larger objects may require more clearance
• Object type: Swimmers and debris have higher priority

Image Processing Settings

Nemo includes advanced image processing to optimize detection in various lighting and weather conditions.

CLAHE (Contrast Limited Adaptive Histogram Equalization)

CLAHE improves local contrast in images, making objects more visible in challenging lighting conditions.

Contrast Adjustment

Manual contrast controls for fine-tuning image appearance.

Gamma Correction

Adaptive gamma correction improves visibility in low-light conditions. This is handled automatically in Lightweight Mode or through the full preprocessing pipeline.

Performance Optimizations

These settings allow you to balance detection accuracy against processing speed.

Lightweight Mode

When to disable: In extremely challenging visibility conditions (heavy fog, dusk/dawn transitions) where full preprocessing may improve detection.

Two-Stage Inference

When to enable Two-Stage:

• When detecting distant objects is critical
• In open water where objects appear small on the horizon
• When FPS is consistently above 45 and you have processing headroom

Detection Sensitivity

Adjust sensitivity thresholds for each object type to fine-tune detection behavior.

Sensitivity Sliders

Each object type has an independent sensitivity slider (0-100%):

Understanding Sensitivity

Lower values (0-25%): More detections, including uncertain ones. May increase false positives but catches more objects.

Higher values (75-100%): Only high-confidence detections. Fewer false positives but may miss some objects.

Recommended Settings by Environment

Video Input & Camera Setup

Nemo accepts virtually any video source. If it outputs video, Nemo can process it.

Supported Video Sources

Camera Positioning

• Height: Mount camera at helm height or higher for best field of view
• Angle: Point slightly downward to capture waterline and horizon
• Coverage: Forward-facing primary, consider additional cameras for 360° coverage
• Protection: Use marine-rated housing for weather protection

Adding an IP Camera or Stream

1. Open Settings > Camera Configuration
2. Select "Add Camera"
3. Enter the stream URL: rtsp://[username:password@]ip_address:port/stream
4. HTTP and MJPEG streams are also supported — paste any valid video URL
5. Test connection and save

Multi-Camera Setup

Nemo supports multiple video sources for comprehensive coverage:

• Each source runs independent detection
• Warnings are consolidated into a single alert system
• Switch between feeds in the dashboard

Streaming Detection Output

Nemo outputs the processed detection feed as a network stream. You can view it in the web dashboard, or consume it from any external application:

• Web browser: Navigate to http://nemo.local:5001
• VLC: Open Network Stream and paste the Nemo stream URL
• OBS / other apps: Add as a media/network source using the stream URL

To get the stream URL, right-click the video feed in the dashboard and select "Copy stream URL" (or similar depending on browser). This lets you display the detection feed on any screen, device, or recording application on your network.

Audio Alerts

Nemo provides audible warnings so you don't have to constantly watch the screen.

Alert Types

Audio Settings

• Volume: Adjust master volume for all alerts
• Alert Types: Enable/disable specific alert sounds
• Repeat Interval: How often collision warnings repeat
• Quiet Hours: Reduce or mute non-critical alerts during specified times

Multi-Device Access

Access Nemo's dashboard from any device on your local network.

Connecting from Other Devices

1. Ensure your device is on the same network as the Nemo Computer
2. Open a web browser
3. Navigate to http://nemo.local:5001 or http://[nemo-ip]:5001
4. The full dashboard is available on any device

Supported Devices

• Smartphones (iOS, Android)
• Tablets
• Laptops
• Chart plotters with web browser capability

Network Requirements

• WiFi or wired network connection to the computer running Nemo
• No internet required — all communication is local
• If your boat doesn't have WiFi, turn on your laptop's Mobile Hotspot (Windows Settings > Network & Internet > Mobile Hotspot) so other devices can connect

Home Assistant Integration

Integrate Nemo with your Home Assistant setup for advanced automation.

Capabilities

• Receive collision alerts as Home Assistant notifications
• Trigger automations based on detection events
• Monitor system status from Home Assistant dashboard
• Control vessel lighting, horn, or other systems in response to alerts

Setup

1. Install the Nemo integration from HACS or manually
2. In Home Assistant, go to Settings > Integrations > Add Integration
3. Search for "Nemo" and select it
4. Enter the Nemo Computer's IP address
5. Configure which entities to expose

Available Entities

Example Automation

automation:
  - alias: "Nemo Collision Alert"
    trigger:
      - platform: state
        entity_id: binary_sensor.nemo_collision_warning
        to: "on"
    action:
      - service: notify.mobile_app
        data:
          message: "Collision warning detected!"
      - service: light.turn_on
        entity_id: light.deck_lights
        data:
          flash: short

Troubleshooting

Common Issues

System Logs

Access detailed system logs at Settings > System > Logs for advanced troubleshooting.

API Reference

Nemo exposes a REST API on the local network, allowing you to control the detection system, read status, and stream video from any device or integration. All endpoints are served from http://<nemo-ip>:5001.

Detection Control

Start Detection

Starts the detection loop. Optionally pass a video_source (camera index or RTSP URL) to switch the input before starting.

curl -X POST http://192.168.1.100:5001/start \
  -H "Content-Type: application/json" \
  -d '{"video_source": 0}'

Stop Detection

Stops the detection loop.

curl -X POST http://192.168.1.100:5001/stop

System Status

Get Status

Returns the current system state including detection counts, FPS, and active collision risks.

// Example response
{
  "running": true,
  "device": "CUDAExecutionProvider",
  "detections": 3,
  "risks": 1,
  "fps": "26.4",
  "total_reflections_filtered": 12,
  "detections_list": [...]
}

Configuration

Get Configuration

Returns the full configuration object including confidence thresholds, preprocessing flags, and risk parameters.

curl http://192.168.1.100:5001/config

Update Configuration

Updates one or more configuration values at runtime. Only include the keys you want to change.

curl -X POST http://192.168.1.100:5001/config \
  -H "Content-Type: application/json" \
  -d '{"lightweight_mode": true, "horizon_position": 0.3}'

Video Streams

Live Video Feed

MJPEG stream of the live detection feed with bounding boxes and risk overlays drawn on each frame. Embed directly in an <img> tag or consume programmatically.

<img src="http://192.168.1.100:5001/video_feed" />

Enhanced Video Feed

MJPEG stream showing the preprocessed (enhanced) frame before detection. Useful for comparing raw vs. enhanced image quality.

Camera Management

List Cameras

// Example response
{
  "camera_sources": ["0", "rtsp://192.168.1.50:554/stream"],
  "active_camera_index": 0
}

Update Camera Sources

Sets the list of available camera sources (USB indices or RTSP/HTTP URLs).

Switch Camera

Switches to a different camera from the configured sources. If detection is running, it will restart automatically with the new camera.

curl -X POST http://192.168.1.100:5001/switch_camera \
  -H "Content-Type: application/json" \
  -d '{"camera_index": 1}'

Debugging & Testing

Test Alarm

Triggers the audio alarm manually. Use this to verify that the speaker and alarm audio are working correctly.

curl -X POST http://192.168.1.100:5001/test_alarm

Debug Mode

Get or set debug mode. When enabled, additional diagnostic information is displayed on the video feed overlay.

curl -X POST http://192.168.1.100:5001/debug \
  -H "Content-Type: application/json" \
  -d '{"enabled": true}'

Quick Reference

Technical Specifications

Supported Hardware

Hardware (Nemo Computer)

Software

Connectivity