zephyr is unable to proccess all my photos

[Hs.Panda]

i'm trying to do a model of my friend and followed the tutorials on how to take the best pictures, but i'm not sure what i'm doing wrong?
i took several photos of her on a white background (i used my door) and had her rotate between each shot (because i wasn't able to move the camera very much)
in total i got 52 photos, from 3 different angles: 1 straight on, 1 looking up and 1 looking down, but zephyr can only proccess about 10 of the photos, even on exhuastive setting.
any help would be appreciated

edit: is it possible that since i'm not moving my camera this could possibly be making zephyr think my all my shots are from the same angle/location? if so is it possible to manually position the photo locations in relation to each other and then have zephyr try to render a cloud from this?

[AvionAvionics]

Hs.Panda Hello Panda, I am just a user learning and getting used to the software as well. But may I make a couple of recommendations. 1. I think the white background may be part of the issue, the more feature zephyr has to pick up on the better. In my personal experience Zephyr has given me fits with items (main and background) that is purely white or black. I have had good success getting individual models like parts and cars modeling by doing rotations around the objects. The moving around the objects seems to work great because it seems to use the background objects as reference points. To my knowledge there isn't a way to add in cameras manually, but I could be quite mistaken on that. Also it may be handy to the developers if you could provide the dataset so they may have an opportunity to take a look.

[Roberto]

Hi Hs. Panda,

As suggested by Avion Avionics, it's better to rotate around your subject. If you keep the camera fixed and rotate the subject, please consider using masking and masquerade to mask out the background. That said, reconstructing a human body with a single camera is not trivial (although not impossible). The main problem is that humans, unfortunately, tend to move . Additionally, interior illumination is usually worse than natural lighting. If possible try to go outside and take photos while walking around her. As a first attempt you can try to focus only on the facial reconstruction, and when you become more expert, you can try to acquire the full body. And if you want to share the datasets with us will be happy to have a look!

[Hs.Panda]

i'll try outside in daylight in the case, thanks for the tips.

here's a few of the photos from the data set i was working with. i'll also try taking new photos

[AvionAvionics]

Hey Panda, If possible do it outside when it is cloudy outside. The darkness from the clouds does wonders. Something to look out for as well is the lighting seems difference in the last 2 than the 1st three. If i am not mistaken that can mess with the alignment as well (the 1st three have a whitish background while the last two have an orange-ish look due to the light..

[Andrea Alessi]

Hi Hs.Panda,

if you zoom in your photos you'll notice that your friend is out focus, while the door is in focus this is going to be a big problem for zephyr, and most likely will try to reconstruct the door rather than your friend.

As previously suggested, try going outside for better lighting, but most importantly make sure that the focus is on the subject and not on the background! If your friend tends to move a lot, you may also consider using a video since it takes less time to shoot a video rather than taking pictures, you can find a nice tutorial here https://3dscanexpert.com/free-3d-sca...eo-smartphone/

I hope this helps!

[Hs.Panda]

thanks for the advice, my friends been sick so i kinda put this on the back burner but finally got a chance to try out your tips, they helped a lot thanks

[Roberto]

Sweet, thanks for the update!