How to Dabarti capture
It’s been a long time since I planned to create this article. And now, since kids are on their vacation, it’s finally a good time.
This article is not sponsored in any form, the author purchased all software and gear.
Probably all of you have seen astonishing foliage renders by Tomasz Wyszolmirski from Dabarti. And some of you may wonder where did he get these beautiful textures. Well, he wrote a small but mighty program that does only one thing, and does it pretty damn good! Extracting real surface normal texture from surfaces.
You can check out more on official Dabarti Capture website HERE.
Of course, nothing good in life is easy, the software has it’s own limitations and flaws, and this little guide will show you how I use it. So. Let’s get started.
Before you start.
- Make sure that what you want to scan, can be scanned by Dabarti capture. It’s not 3d scanning software. It can extract depth from flat surfaces like fabrics, wood, skin, and more. But it can’t create 3d surfaces clothes, boxes or hands.
- Prepare some safe, calm environment, without running animals and other people that can potentially interrupt the scanning process.
This is gear I’m using on my workflow.
- Multifunction remote A – For triggering a camera remotely.
- Multifunction remote B – For triggering camera and flash.
- Shutter Release Cable
- Speedlight with built-in receiver.
- Tons of good quality rechargeable batteries. – I’m using the gold standard. Eneloops from Panasonic.
- Camera – I’m using Nikon z7 with Sigma 85mm Art lens attached.
- Simple brush for dust cleaning.
- Lens polarisation filter.
- Polarisation foil.
- Colour reference chart. – I’m using Datacolor SpyderCHECKR
- Chrome ball – Everything that’s perfectly round will works. I’m using a small door handle.
- Tripod with the multi-angle centre column. – I’m using one of Manfrotto’s.
- I placed the camera on a steady tripod, framed the shot and tighten all joints.
- I turned ON camera flash through the hot shoe.
- Multifunction remote A was set as camera release transmitter.
- Multifunction remote B was set as camera release receiver and flash transmitter at once. It was connected into camera’s release port so it can release a shutter, and it was installed into camera hot shoe so it can release the flash.
- Speedlight was set to receive a fire signal from Multifunction remote B.
- Polarisation filter was placed on the camera lens.
- Polarisation foil was sticked to speedlight.
- I synchronised polarisation. (More on that later.)
- All camera parameters were set to Manual:
a. ISO 100.
b. Shutter speed 1/100.
c. Aperture F9 for good optical performance and wide focus planes.
d. White balance fixed at 5300k (Exact color temperature is not so important since we will calibrate it later.)
e. Image quality Full Frame 14-bit RAW lossless.
f. I turned off all image processing filters.
g. Color profile Standard.
h. I turned off applying settings for a live preview for easy framing.
- I turned off manual focus and focused image manually in the live view zoom window.
- I placed a chrome ball in the top right corner of the frame.
- I made a reference photo of the color checker. If you don’t have one, simply made a photo of a plain sheet of paper. It will be much easier to catch the correct white balance in Lightroom.
Light polarisation is not required, but since it’s so effective, it is highly recommended to do so. By using polarised light, you can illuminate shiny surfaces without signs of specular light. This works like magic.
Capturing itself is the most straightforward part of the process. I was triggering multifunction remote A from one hand, and moving speedlight to different positions from another hand. I made around 15 photos with varying positions of light. (5 to 8 photos would be enough, but I’d like to have some good light coverage). There are two essential things in this step.
- Since all captured images have to be perfectly aligned, be careful, not to move your camera, tripod, chrome ball or subject.
- To take advantage of light polarisation, point your Speedlight correctly. Keep 90° between your lens polarisation filter and Speedlight polarisation filter.
- After taking a series of photos, I imported all images into Lightroom.
- I made a color profile from color checker. If you don’t have one, simply pick white balance from a sheet of paper you photographed instead.
- All images were exported in full resolution TIF’s for processing.
- It’s essential to prepare the black-white mask for the probe ball. I made one quickly in photoshop. Remember to name it correctly “probe_mask.jpg”.
Since Dabarti Capture doesn’t use a graphic user interface, the way of triggering the whole process is a bit odd.
There are two ways to run Dabarti Capture.
First of all, all TIF images and probe_mask.jpg have to be saved in one folder.
Drag project folder icon on “capture.exe” file. The program will automatically recognise files and start to solve light.
Create “work.bat” file in notepad with commands for the program. This method allows you to process multiple image sets automatically. Specific process and controls are explained in the manual.
The whole process takes some time. It depends on image resolution and hardware.
After the process is complete, several maps are saved in the Output folder.
Unfortunately, Albedo map looks ugly and I’d like to make one myself. I’m just blending all 15 images with “Lighten” mode.
Since the whole process is not so fast and easy, the results are mind-blowing. Software managed to reproduce the smallest details.
Things to remember.
- Use the right size reflective ball for the job. Give to software something to calibrate.
- Fix camera rocksteady and release the shutter remotely. Even the slightest shift between photos will cause errors.
- Everything on manual!
- Keep the right angle between polarisation filters.
- Take more photos than you need. This will eliminate shadowed areas and let the software for better computation.
The software already had few updates, but there’s always fields for improvement. Here are my suggestions:
- Graphic user interface
- Albedo produced by the software is often too dark.
- Depth solving requires a lot of RAM
- Better multi core support.