My $180 DIY Home Security Camera System (No Monthly Fees, Local Storage Only)

Why I Ditched Ring

I used Ring cameras for 2 years. Four cameras: doorbell, garage, backyard, front yard.

What I paid:

• Cameras: $600 total (4× $150 each)
• Ring Protect subscription: $200/year (required for video history)
• 2-year total: $600 + $400 = $1,000

What I got:

• Constant privacy concerns (Amazon owns Ring, uploads everything to AWS)
• 60-day video history max (can’t go back further)
• Slow notifications (5-10 second delay)
• Frequent false alerts (trees blowing = “person detected”)
• Cloud dependency (internet down = no recordings)

I wanted:

• Local storage (no cloud uploads)
• Person detection (ignore trees, cars, animals)
• Unlimited history (30+ days of 24/7 recording)
• No subscription fees (one-time cost only)

Enter: Frigate NVR (Network Video Recorder).

Ring uploads everything to Amazon’s cloud. Frigate keeps everything local on your NAS.

What Is Frigate NVR?

Frigate is open-source security camera software.

Key features:

• AI object detection: Detects people, cars, pets (ignores trees, shadows)
• Local processing: Everything runs on your hardware
• RTSP support: Works with any IP camera (not locked to one brand)
• 24/7 recording: Limited only by your storage space
• No subscription: Free forever (donate if you like it)

Frigate runs on:

• Raspberry Pi (budget option)
• Old laptop/desktop (easy)
• Docker on NAS (what I use)
• Dedicated server (overkill for home)

I run Frigate on my TrueNAS server (built in a previous project). Total cost: $0 (already owned hardware).

Hardware: Cameras + PoE Switch

Cameras: Reolink RLC-810A (4× $45 = $180)

Reolink RLC-810A is a PoE (Power over Ethernet) camera.

Specs:

• Resolution: 4K (3840×2160)
• Night vision: IR LEDs, 100ft range
• Power: PoE (powered by Ethernet cable, no separate power adapter)
• Protocols: RTSP, ONVIF (works with any NVR)
• Storage: Supports microSD + NVR
• Price: $45 each (4-pack = $180)

Why PoE cameras?

PoE is the only serious option for 24/7 recording. Wi-Fi cameras drop frames. Battery cameras sleep (miss events).

Reolink vs competitors:

Reolink wins on price and compatibility. Hikvision has better AI but costs 3× more (and Frigate’s AI is better anyway).

PoE Switch: TP-Link TL-SG1008P ($80)

TP-Link 8-port PoE switch powers cameras and connects them to network.

Specs:

• 8 ports: 4× PoE ports + 4× regular Ethernet
• PoE budget: 65W total (cameras use ~10W each)
• Gigabit: 1 Gbps per port
• Fanless: Silent operation

Why this switch?

• 4 PoE ports: Perfect for 4 cameras
• 65W budget: 4 cameras × 10W = 40W (25W headroom)
• Fanless: Silent (important for home office)
• Cheap: $80 (managed PoE switches cost $200+)

Wiring:

1. Ethernet from router to switch (uplink)
2. Ethernet from switch to each camera (PoE delivers power + data)

One cable per camera. No power adapters needed.

PoE switch setup: Router → Switch → 4 cameras (one cable each)

Storage: WD Purple 2TB ($60)

WD Purple 2TB drive is designed for surveillance (24/7 writes).

Why surveillance drives?

Surveillance drives handle constant writes better. Desktop drives fail within months under 24/7 recording.

Capacity calculation:

4 cameras × 4K resolution × 15 fps × 24 hours = ~200 GB/day

2 TB = 10 days of 24/7 recording (before overwriting oldest footage)

I keep 30 days of “events” (motion detected) and 10 days of continuous recording.

AI Accelerator: Google Coral USB ($60 optional)

Frigate uses AI to detect people/cars/pets. This requires processing power.

Options:

• CPU only: Works, but slow (5-10 fps). Maxes out CPU.
• GPU (NVIDIA): Fast (30+ fps), but requires beefy GPU ($300+).
• Google Coral USB: Purpose-built AI chip. 100+ fps. Costs $60.

I bought the Google Coral USB Accelerator ($60). Plug-and-play, runs cool, uses 2W.

Result: CPU usage dropped from 80% (software AI) to 15% (Coral offload).

Installation: Running Ethernet Cables

This was the hardest part. Running 4× Ethernet cables from my office (where NAS + switch live) to camera locations.

Option 1: Attic/Crawlspace (What I Did)

Process:

1. Drill hole from attic into wall cavity
2. Fish cable down wall cavity to camera location
3. Drill hole through exterior wall
4. Mount camera, plug in Ethernet

Tools needed:

• Drill + 1” spade bit
• Fish tape (for pulling cables through walls)
• Ethernet crimping tool (for custom cable lengths)
• Ladder

Time: 6 hours for 4 cameras (1.5 hours per camera)

Difficulty: Hard if you’ve never fished cables. Watch YouTube tutorials first.

Option 2: External Conduit (Easier but Ugly)

Run cables outside the house in weatherproof conduit.

Pros: No attic/wall access needed Cons: Visible cables on exterior (ugly)

Option 3: Hire an Electrician ($300-500)

If you’re not comfortable with attic work, hire a low-voltage electrician.

Cost: $75-125 per cable drop × 4 cameras = $300-500

I saved $500 by DIY, but it took 6 hours.

Fishing Ethernet cable from attic down to exterior camera location

Frigate Setup: Docker on TrueNAS

I run Frigate in a Docker container on my TrueNAS server.

Step 1: Install Docker

TrueNAS SCALE has built-in Docker support (called “Apps”).

Apps → Available Applications → Search “Frigate” → Install

Step 2: Configure Cameras

Frigate uses a YAML config file to define cameras.

Example config (/config/config.yml):

mqtt:
  enabled: False

detectors:
  coral:
    type: edgetpu
    device: usb

cameras:
  front_door:
    ffmpeg:
      inputs:
        - path: rtsp://admin:password@192.168.1.101:554/stream
          roles:
            - detect
            - record
    detect:
      width: 1280
      height: 720
      fps: 5
    objects:
      track:
        - person
        - car
    record:
      enabled: True
      retain:
        days: 10
    snapshots:
      enabled: True
      retain:
        default: 30

Repeat for all 4 cameras (front_door, backyard, garage, driveway).

Step 3: Mount Storage

Frigate stores recordings in /media/frigate/recordings.

I created a dataset on my NAS: /tank/frigate (500 GB quota).

Docker volume mount: /tank/frigate:/media/frigate

Step 4: Plug in Google Coral

USB Coral plugs into NAS. Docker container accesses it via USB passthrough.

Config: device: usb (Frigate auto-detects Coral)

Step 5: Start Frigate

Apps → Frigate → Start

Frigate boots in 30 seconds. Web UI: http://nas-ip:5000

Frigate web interface: 4 live camera feeds + event timeline

AI Detection: Filtering Out False Positives

Ring sent me 30+ notifications/day. 90% were false positives (trees, shadows, cars driving by).

Frigate’s AI filters this:

Detection config:

objects:
  track:
    - person
    - car
  filters:
    person:
      min_area: 5000  # Ignore tiny detections (distant people)
      threshold: 0.75  # 75% confidence required
    car:
      min_area: 20000
      threshold: 0.70

Result:

• Ring: 30 alerts/day (3 relevant, 27 false)
• Frigate: 4 alerts/day (4 relevant, 0 false)

What Frigate ignores:

• Trees swaying in wind
• Shadows moving across driveway
• Cars driving past (on street, not driveway)
• Small animals (squirrels, birds)

What Frigate detects:

• Person walking up to front door
• Car pulling into driveway
• Package delivery

Notifications are now useful instead of annoying.

Performance: CPU and Storage Usage

CPU Usage

Before Google Coral: 80% CPU (TensorFlow running on CPU) After Google Coral: 15% CPU (AI offloaded to Coral)

The Coral is mandatory if you want smooth performance.

Storage Usage

4 cameras × 4K × 15 fps × H.264 = ~200 GB/day

Retention policy:

• 24/7 recording: 10 days (200 GB × 10 = 2 TB)
• Event clips: 30 days (5 GB)

Total: 2 TB drive is perfectly sized.

Frigate auto-deletes oldest footage when disk is full. No manual cleanup needed.

Network Bandwidth

4 cameras × 4K × 15 fps = ~80 Mbps sustained

My gigabit network handles this easily. PoE ensures reliable connection (no Wi-Fi dropouts).

Mobile Access: Frigate App + VPN

Frigate has iOS/Android apps. But I don’t expose Frigate to the internet (security risk).

Instead, I use WireGuard VPN:

1. Connect to VPN when away from home
2. Access Frigate web UI: http://nas-ip:5000
3. View live cameras + recorded events

Alternative: Reverse proxy with SSL (more complex, not worth it for home use).

Cost Breakdown: $380 Total (No Monthly Fees)

Compare to Ring:

• Year 1: $600 (cameras) + $200 (subscription) = $800
• Year 2: $200 (subscription)
• Year 3: $200 (subscription)
• 3-year total: $1,200

My system:

• Year 1: $440 (one-time)
• Year 2: $0
• Year 3: $0
• 3-year total: $440

Savings: $760 over 3 years

And I get:

• Better AI detection (fewer false positives)
• Unlimited local storage (not 60-day cloud limit)
• No privacy concerns (nothing uploaded to cloud)
• Faster notifications (local processing = instant alerts)

Real-World Testing: Detection Accuracy

I tracked detection accuracy for 30 days.

Person Detection

Total person events: 47 Correctly detected: 46 (97.9%) False positives: 3 (tree branch shadow looked like person) False negatives: 1 (missed one delivery driver, was 100+ feet away)

Verdict: Excellent. Better than Ring (which flagged cars as “person”).

Car Detection

Total car events: 23 Correctly detected: 23 (100%) False positives: 0 False negatives: 0

Verdict: Perfect. Only triggers when car enters driveway (ignores street traffic).

Pet Detection (Optional)

I enabled pet detection for backyard camera.

Total pet events: 18 (neighbors cat + local squirrels) Correctly detected: 16 (88.9%) False positives: 2 (shadow on fence = “dog”)

Verdict: Good enough. I don’t care about every squirrel.

Power Consumption: 35W Total

Measured with Kill-A-Watt:

Wait, that’s 59W, not 35W?

The NAS was already running 24/7 (18W). Frigate adds negligible CPU load thanks to Coral. I’m only counting the new components.

Cost per year: 59W × 24h × 365d ÷ 1000 × $0.12/kWh = $62/year

Compare to Ring:

• Ring cameras: 8W total (battery charging)
• Cost: $8.40/year
• But: $200/year subscription

My system: $62/year electricity, $0 subscription.

Still $138/year cheaper than Ring.

Issues and Limitations

1. Complex Setup (Not For Everyone)

Ring is plug-and-play. Frigate requires:

• Running Ethernet cables (6 hours DIY or $300-500 hire electrician)
• Setting up Docker container
• Configuring YAML files
• Understanding networking (RTSP, VLANs, VPN)

Not beginner-friendly.

If you want simplicity, stick with Ring/Wyze.

2. No Cloud Backup (By Design)

All footage is local. If someone steals my NAS, I lose all recordings.

Mitigation:

• NAS is hidden in locked closet
• Frigate uploads “events” (person detected) to cloud storage (optional)

For me, privacy > cloud backup. But this is a trade-off.

3. Google Coral Is Hard to Find

Google Coral USB was frequently out of stock (supply chain issues). I waited 2 weeks for restock.

Alternative: Use a cheap NVIDIA GPU (GTX 1050 Ti = $100 used). Requires more power but works great.

Should You Build or Buy Ring?

✅ Build Frigate if:

• You value privacy (no cloud uploads)
• You’re comfortable with Linux/Docker/networking
• You want unlimited local storage
• You hate subscription fees
• You already own a NAS or old PC

❌ Buy Ring/Wyze if:

• You want plug-and-play simplicity
• You don’t care about cloud storage
• You don’t mind $200/year subscription
• You rent (can’t run Ethernet cables)

Verdict: Best Security Investment I’ve Made

Frigate NVR transformed my home security.

Before (Ring): Cloud dependency, constant false alerts, $200/year fees, privacy concerns. After (Frigate): Local control, accurate AI detection, $0 ongoing costs, total privacy.

$440 upfront, $0/year ongoing. Saves $200/year vs Ring.

Payback period: 2.2 years. After that, pure savings.

Rating: 5/5

Recommended? If you’re tech-savvy and value privacy, absolutely build this.

Next Steps: Expanding the System

Now that my security cameras are dialed in, I’m exploring:

• License plate recognition (Frigate supports this)
• Home Assistant integration (automate lights when person detected)
• Facial recognition (recognize family vs strangers)
• Offsite backup (upload events to Backblaze for redundancy)

The self-hosted rabbit hole continues.

Setup time: 8 hours (6 hours cable running, 2 hours software setup) Last updated: September 19, 2025