Getting Started - How to Use Calibrator
You’ve installed your new sensor, you’ve made sure everything works, hardware-wise. The last thing you need to do before scanning is to calibrate. Using Calibrator to calibrate your sensor’s depth feed with the iPad’s color feed is far from intuitive (though we are working to improve this experience every day).
This guide is intended to be used alongside this tutorial video as a complete guide to successfully calibrating (which we promise, can be done!).
- What is the importance of calibration?
- Step 1: Wide Vision Lens Calibration
- Step 2: Outdoor Calibration Mode
- Step 2: Indoor Calibration Mode
- Step 3: Refinement
What is the importance of calibration?
Both the original Structure Sensor and Structure Sensor (Mark II) work in tandem with your iPad’s color camera. The iPad’s camera applies the high-definition color and texture of the objects being scanned; Structure Sensor captures depth.
The problem is, the two cameras exist in two different places, and if run without calibration, would produce effects where the color from one object may shift over into the color of another. For example, say we were to scan this couch:
If we scanned the couch without calibrating the sensor properly, the scan might result in something like this:
See how the blanket bleeds into the pillow? That’s not right.
So we need calibration to inform the sensor and software of its place in space in relation to the iPad’s camera. When properly calibrated, the scan looks something like this:
As we can see here, there is clear separation between the blanket and the pillow, and the color lays naturally where it should.
The sensor stores its calibration values locally, and it associates them with the iPad that was used to calibrate. Once you calibrate properly, you will not need to do so for quite a while longer, if ever.
The sensor holds two states of calibration: with and without the Wide Vision Lens.
Step 1: Wide Vision Lens Calibration
Things get even trickier with the addition of the Wide Vision Lens, as the fisheye balloons out the color stream. To help calibrate with the Wide Vision Lens, Calibrator introduces an extra step, utilizing this checkerboard.
- To use the checkerboard, print it out in landscape and make sure the squares are exactly 25mm in each direction.
- Post the checkerboard on a flat surface and make sure there are no bubbles.
- Attach your Wide Vision Lens, Launch Calibrator, and tap “Calibrate with Wide Vision Lens” at the bottom.
- Read through the tutorial and tap “Start Calibration”.
- For the next part, you will need to position the iPad in various angles, matching the app’s AR overlay with the printed checkerboard. This can be a little confusing to start.
- The easiest way to think about how to position the iPad is to try and make sure the top right corner of the overlay matches up with the checkerboard, and then adjust the iPad’s angle to match the rest.
- The angles from which you will need to approach the checkerboard are as follows:
|Straight on, closer, and slightly from the bottom|
|From the right center, with the left edge of the iPad angled evenly away from the checkerboard|
|From the right top, with the left bottom edge of the iPad angled evenly away from the checkerboard|
|From the center top, with the bottom edge of the iPad angled evenly away from the checkerboard|
|From the left top, with the right bottom edge of the iPad angled evenly away from the checkerboard.|
Step 2: Outdoor Calibration Mode
Both calibration states (with and without the Wide Vision Lens) require bracket calibration. This is the software alignment between the iPad’s camera and Structure Sensor.
Indoor and Outdoor Calibration modes are fundamentally the same; what changes is the gain and exposure settings of the sensor. Typically we suggest performing Outdoor Calibration, but if you live in a particularly cloudy area or are having trouble, we provide Indoor Calibration as an alternative.
- To begin bracket calibration, tap “Start Calibration”.
- Depending on your iOS device, you may be taken to a screen that discusses your bracket type. For all supported iPads, we include the XYZ extrinsic translations by default--these are the distance between the sensor and the iPad’s camera in terms of height, width, and depth in millimeters. If you have a custom case, you must make these measurements yourself (if you need to make these measurements, please check out this page). Assuming you have a bracket, tap your bracket type.
- You will be taken to a screen that says “Sunlight Required”.
Fill up the sun meter by pointing the sensor outside or through an open window. If there is not enough IR available, you will see this toggle appear after five seconds:
At this point, you can either tap the toggle to adjust the exposure and gain to begin outdoor calibration, or continue pointing your sensor outside to try and fill the meter.
- Once full, you will see a split screen of color and IR. This is where bracket calibration occurs.
Your goal here is to find a complex scene that is high enough in contrast that it appears well in both screens. In the picture below, I use the bare trees against the sky:
As you can see, the trees show up fairly clearly in the IR feed as well as the color feed. This is an example of a scene with poor complexity.
- Once you find a complex enough scene, you don’t need to move a huge amount between frames--just move a little! This way you don’t have to continually find complex enough scenes.
- After two or three good frames, the app will switch to the final stage: refinement.
Step 2: Indoor Calibration Mode
If you are calibrating at night, or in an area of the world that doesn’t see a lot of clear skies, you may not have enough ambient IR to complete Outdoor Calibration. Not to worry! If you are at this screen for five seconds, a toggle in the lower left corner will show up to enable Indoor Mode.
What this does is adjust the exposure and gain settings of the sensor, making it more sensitive to IR light.
The rules of calibration remain the same as Outdoor Calibration Mode; you need to ensure you are scanning a complex enough scene. What constitutes a complex indoor scene? A scene that has a lot of contrast on both the IR stream as well as the color stream. An easy example is a MacBook keyboard:
This is because the black keys are clearly contrasting with the light gray chassis.
You might imagine a plant would be a complex enough scene. Not so!
You might also imagine a geometrically complex scene would be complex enough. Again, not necessarily!
In this picture, we see a geometrically complex scene, but there are very little high contrast areas in the IR stream.
While not perfect, the above scene was enough to successfully complete bracket calibration. The high contrast between dark and light created enough similarity between the scenes, and the software was able to do the rest.
If you have trouble finding a complex enough scene, try downloading and printing this convenient indoor calibration target. It turns out Jackson Pollock knew a thing or two about complex scenes.
Step 3: Refinement
If you’ve made it this far, congratulations! You’ve made it through the hardest part.
Refinement is the last part of calibration, where you manually tweak the calibration the software determined was correct. We do this by allowing you to drag the depth feed’s color overlay to match the physical scene.
Please note! You do not need to perform refinement while looking outside! In fact, you might have better luck doing so indoors. We do our best to help you get the hang of it by working through this tutorial:
Once you think you’ve got it, find a scene with some very clearly-defined edges, like this scene here:
On iPads, you will drag the scene right and left. On iPhones, you will drag the scene up and down.
Once the color (not the spaces where there is no color!) matches the physical scene, tap “Save Calibration”.
The calibration state will save to the sensor.
Remember, there are two calibration states that need to be saved: with and without Wide Vision Lens, so be sure to calibrate with both.
Once you’re fully calibrated, you are good to go.