Problem with taking the right photos for reconstrucion

**Andrea Alessi** · 2017-12-01, 04:05 PM

Hi szafir,

1. in the first set (the black car) i would also mask the images, as there aren't many features on the car itself, so zephyr is also picking up features from the wall. Luckily, since there is a very high contrast you can easily make masks in few clicks, in masquerade use the red brush on the center of the subject, use the blue brush on the wall, and use the autocompute function.

2. the second set is harder, as many images are out of focus ( 20171130-PB300671.jpg ) and other are underexposed. You should also mask this one due to the reflections on the table. I'm not sure how well this dataset would perform however, as the images aren't very sharp and the subject isn't very textured.

3. the third one is also very hard, as it is not feature rich enough, glossy, and there are few photos. As a workaround, you should try and dirty the surface, use a pico projector or maybe try with a newspaper below it to help the first phase. Anyways, you should take more pictures.

When dealing with close range, i also suggest you use a different aperture, so that everything is in focus. Also avoid flash, try to use diffused lighting.

You can find some tips on photography and an example of the projector trick on this videotutorial https://www.youtube.com/watch?v=E06kgYBftak

Unfortunately i can not process them now as all our workstations are busy, but i'll try to do that in the next few days, but i would expect Zephyr to discard most images

**szafir51** · 2017-12-02, 08:30 AM

Thank you for your valuable suggestions. I'll try to prepare new set of photos in daylight. I'll also paint the car body with different colours.

Problem with taking the right photos for reconstrucion

Problem with taking the right photos for reconstrucion

Comment

Comment