Skip to main content

Troubleshooting & FAQ (OV20i)

Reference page, browse as needed

Stuck? Start here. This page covers the most common issues customers hit with the OV20i, organized by the step where they happen.

Installation issues​

Camera won't power on​

Power troubleshooting table
CheckFix
Power supply voltageMust be 19-24 VDC. Outside this range = no boot
Minimum currentPower supply must provide at least 1A (18W max)
Cable connection17-pin M12 power connector must click and lock. A loose connection looks like "no power"
Power pinsPins 13 or 14 = 24V DC (+), Pins 5 or 6 = GND
LED behaviorNo LEDs at all? Check the power source with a multimeter. If LEDs flash briefly then die, the supply can't sustain the load
Wall power?

You cannot plug the camera into a standard wall outlet. It requires industrial 19-24 VDC power. Use the included power supply or an equivalent industrial supply.

Image is blurry​

  1. Check focus. Adjust the software focus slider in the browser UI until the image sharpens. Use the live preview to see changes in real time
  2. Check working distance. Each lens has an optimal focus range. If the camera is too close or too far, the lens physically can't focus
  3. Check for vibration. If the image is sharp when the line is stopped but blurry during production, the camera or part is vibrating. Tighten the mount or reduce conveyor speed
  4. Check exposure. An exposure time that's too long + moving parts = motion blur. Lower the exposure and compensate with more light or higher gain

Image is too dark or too bright​

  • Too dark: Increase exposure time, increase gain, increase built-in LED intensity, or add external lighting
  • Too bright: Decrease exposure, lower gain, reduce LED intensity
  • Uneven lighting (bright spots and shadows): Adjust the LED light pattern in software. The OV20i's built-in LEDs work well for close-range inspections. For longer working distances, consider adding external lighting

Wrong lens: part doesn't fill the frame​

Lens and field-of-view troubleshooting
ProblemSolution
Part is tiny in the frameUse a longer focal length lens (e.g., 16mm or 25mm) or move the camera closer
Part extends beyond the frameUse a shorter focal length lens (e.g., 6mm or 8mm) or move the camera farther away
Can't get close enoughSwitch to a macro lens configuration

Refer to the Lens + FOV table in the Reference section for exact field-of-view calculations per lens.

Connection issues​

Can't reach the camera IP​

  1. Verify same subnet. Your computer must be on the same network subnet as the camera. Default camera IP is 192.168.0.100, so your computer needs an IP like 192.168.0.50
  2. Ping test. Open a terminal and run ping 192.168.0.100. If you get "Request timed out," there's a network issue
  3. Check cables. Ethernet cable must be securely connected at both ends. Try a different cable
  4. Firewall. Your computer's firewall may be blocking the connection. Temporarily disable it to test
  5. Wrong IP. If someone changed the camera's IP from the default, you'll need to find it. Check your router's DHCP lease table or use a network scanner
  6. Emergency USB. Connect via USB and access the camera at 192.168.55.1

Browser shows a blank page or error​

  • Use Chrome. Google Chrome is the recommended browser. Firefox and Edge work too. Safari has limited compatibility
  • Clear cache. Try Ctrl+Shift+R (hard refresh) or clear your browser cache
  • Wait for boot. The camera takes about 30 seconds to fully boot. If you connected power just moments ago, wait and try again
  • Check the URL. Make sure you're going to http:// (not https://) followed by the camera's IP address

Camera boots but no live image​

Video feed troubleshooting
SymptomLikely cause
Page loads but video feed is blackLens not installed, or something is obstructing the lens
Video feed shows but is completely darkLighting isn't on, or exposure is set too low. Check built-in LED settings
Intermittent video freezingNetwork bandwidth issue; use a direct Ethernet connection, not WiFi
"Camera not found" messageBrowser lost connection. Refresh the page

LED troubleshooting​

All four LEDs on the OV20i are located on top of the camera.

LED stateMeaning
Power LED solid greenPower OK
Power LED offNo power; check cables and supply
System LED slow blinkBooting up; wait 30 seconds
System LED solidCamera ready
System LED fast blinkError; check connections

Recipe issues​

Alignment keeps failing​

This is the #1 support question. Here's a systematic fix:

  1. Check template regions. You need 2-3 small regions placed as far apart as possible on the part. If your regions are too close together or too large, the aligner struggles

  2. Look at the edge highlights. Template regions work on edges (high-contrast boundaries). If there are no strong edges in your region, the aligner has nothing to lock onto. Move the region to a feature with clear edges

  3. Clean up noise. Use the Ignore tool (eraser) to remove edges from things that move or change between captures: conveyor texture, background clutter, reflections

  4. Adjust sensitivity. The sensitivity slider controls how aggressively the aligner looks for edges. Too high = picks up noise. Too low = misses your part. Start at the default and adjust

  5. Check confidence threshold. If the aligner finds the part but reports low confidence, lower the threshold slightly. If it's finding false matches, raise the threshold

  6. Verify physical setup. The aligner handles XY position + rotation, but NOT 3D changes. If your part tilts, flips, or changes perspective between captures, the aligner can't compensate. Fix the physical fixturing

The most common alignment mistake

Placing template regions on things that move or change between captures (like a conveyor belt edge, a shadow, or a flexible cable). The aligner sees these as part of your template and gets confused when they look different next time.

For a full guide, see Step 2: Alignment.

AI model accuracy is poor​

Before retraining, check these first:

  1. Mislabels. Click View All ROIs and verify every single label. One wrong label in a small dataset is catastrophic (1 mislabel out of 5 images = 20% corruption)

  2. ROI too large. If your ROI is bigger than 512x512 pixels, the image gets downscaled and small defects disappear. Make your ROIs smaller and more targeted. See Step 3: Inspection Regions

  3. Image quality. If you can't see the defect clearly in the raw camera image, the AI can't learn it. Go back to lighting and focus

  4. Not enough variety. Your training data should include the range of variation the AI will see in production: different lighting conditions, different part positions, different batches

If those check out, then retrain:

  1. Open the Library, find the images the AI got wrong
  2. Add them to the training set, fix labels if needed
  3. Retrain, and focus on failures, not random new data

ROI shows downscale warning​

If you see a warning that your ROI is being downscaled, your inspection region is too large. The AI processes a maximum of 512 x 512 pixels per ROI. Anything larger gets shrunk to fit, and you lose detail.

Fix: Make the ROI smaller. If you need to cover a large area, split it into multiple smaller ROIs using the grid tool.

Pass/fail output isn't working​

  1. Check the IO Block. Navigate to the IO Block in your recipe editor. Verify the rules are set correctly (e.g., class must equal "pass")
  2. Check trigger mode. Is the camera actually triggering captures? Try manual trigger first
  3. Check wiring. If using digital outputs, verify the M12 I/O cable is connected and wired correctly to your PLC or stack light
  4. Check Node-RED. If using Advanced Mode, open Node-RED and check for errors in the flow. Look at the debug panel for error messages

Frequently asked questions​

Do I need internet?​

No. The camera runs entirely on your local network. You don't need internet for:

  • Configuration and recipe creation
  • Running inspections
  • Viewing results and the Library

Internet is only needed for:

Can I use a different power supply?​

Yes, as long as it meets the specs: 19-24 VDC, minimum 1A (18W max). The camera ships with a suitable power supply, but any industrial supply meeting these specs will work. Do not use a standard USB or wall adapter.

What if I need to inspect multiple part types?​

Create a separate recipe for each part type. You can switch between recipes:

  • Manually through the browser UI
  • Automatically via PLC command
  • Using a barcode scanner + Node-RED flow

How many images do I need to train the AI?​

Start with 3-5 images per class (e.g., 3-5 good, 3-5 bad). Train in development mode, test, and add more images specifically where it fails. Most inspections work well with 5-20 images. Complex multi-defect problems may need 50-500+ images.

The AI has high learning capacity, and it keeps improving with more data, unlike many systems that plateau after 20 images.

Classifier or segmenter?​

If you need...Use a...
Pass/fail decisionClassifier
Presence/absence checkClassifier
Pixel-level defect locationSegmenter
Defect size measurementSegmenter
Multiple defect types in one areaSegmenter

When in doubt, start with a classifier; it's faster to set up. See Classifier vs. Segmenter for the full guide.

How do I reset the camera?​

The camera has a physical reset button. Press and hold for:

  • Short press (< 5 seconds): Restarts the camera (keeps all settings)
  • Long press (> 10 seconds): Factory reset (erases all recipes and settings)
Factory reset is permanent

A factory reset deletes all recipes, training data, and settings. Only use this as a last resort. Download your recipes and logs first if possible.

Who do I call for help?​

  1. AI Assistant at tools.overview.ai. Available 24/7, knows everything about the camera
  2. Support email. Contact Overview.ai support through your account
  3. Learning Center at overview.ai/resources/learning-center. Interactive walkthroughs

Can I access the camera remotely?​

The camera is accessible from any device on the same network via a browser. For remote access from outside the local network, you'd need to configure VPN access or port forwarding on your network (consult your IT team).

Still stuck?​

If the video and this page don't solve your problem, describe your issue to the AI Assistant at tools.overview.ai. It can walk you through advanced diagnostics step by step.