The Waterfall Principle
Time: 2 minutes (reading)
This is the single most important concept to understand before you start. It will save you hours of rework.
The rule
Only move to the next step once you've fully optimized the current step.
Every step in creating a recipe builds on the one before it. If you skip ahead and then need to change something earlier, you have to redo everything that came after.
Physical Setup
Mount · Lens · Light
Image Settings
Exposure · Gain · Lens
Template & Aligner
Capture · Align edges
Inspection Regions
Draw ROIs
AI Training
Label · Train model
Outputs
Pass/Fail · Node-RED
Why it matters: a real example
Imagine you rush through image settings without enabling lens distortion correction. You capture your template image, spend 20 minutes setting up the aligner, carefully draw 15 inspection regions, and start training your AI model.
Then you notice the aligner is unreliable near the edges of the frame. You realize it's because of lens distortion, something you should have fixed in image settings.
Now you have to:
- Go back to image settings and enable lens correction
- Recapture the template image (the old one is now invalid)
- Redo the entire alignment setup
- Reposition all 15 inspection regions
- Retrain your AI model
20 minutes of skipping turned into 2 hours of rework.
If you go back and adjust image settings after training, you will need to recapture the template, redo alignment, reposition all inspection regions, and retrain your AI model. This applies to any setting that changes how the image looks, including exposure, gain, lens correction, and white balance.
The good news
When you follow the waterfall, every step gives you visual feedback. You can always click Live Preview to verify your work before moving on. The camera is designed to make this easy, but only if you go in order.
Think of it like building a house. You wouldn't paint the walls before the foundation is set. Each step in the recipe is a layer. Get it right, verify it, then move on. Customers who understand this upfront find the camera works amazingly well.
The steps in order
| Step | What you do | What to verify before moving on |
|---|---|---|
| 1. Physical Setup | Mount camera, choose lens, set up lighting | Image is sharp, well-lit, part fills the frame |
| 2. Image Settings | Exposure, gain, lens distortion correction | Clean, consistent image in Live Preview |
| 3. Template & Alignment | Capture template, set up aligner | Alignment tracks the part reliably in all positions |
| 4. Inspection Regions | Draw ROIs on areas to inspect | ROIs are small, well-positioned, move with alignment |
| 5. AI Training | Label images, train model | Development mode shows correct results |
| 6. Outputs | Configure pass/fail rules, IO | Global pass/fail matches your expectations |
Use Development Mode to verify each step before moving on. It trains a quick model in about 30 seconds so you can confirm the inspection is working correctly. This is the fastest path to a reliable inspection, catching problems early when they are cheap to fix rather than late when they require a full rework.
Ready to start? Go to Install the Camera.