What lurks in a Staffordshire GardenDaily Photogrammetry Challenge, Week Two

Posted on 10 April, 2020

Doom was the thematic bond between last week’s entries to my Daily Photogrammetry Challenge.

For week two, I’ve chosen something very similar – our garden.

I’ve been outside maybe four times since Coronavirus became a thing people were taking seriously. Obviously by people, I’m not including the inbred troglodytes who are continuing as normal.

This little challenge seemed a good excuse to get myself some fresh air. There’s also plenty of interesting crap in both our front and back gardens.

So, without further adoo…

Story continues below...

Topics: Game Design,Photogrammetry,Reality Capture

Heavy Garden Table

April 3rd 2020

What worked: Overall geometry and texture quality, generally consistent level of detail throughout.

What didn’t work: Finer details on surface (better angles and focus would’ve helped), inconsistent colour (needed to adapt camera settings for each angle), underside (couldn’t get low enough).

Photocount – 271 (252 accepted)
Photo positions/angles – 7
Photo time – 15 minutes
Post processing – None, lazy
Total reconstruction time – 1.5 hours
High detail polycount – 44.5M tris

Bird Table

April 4th 2020

What worked: Consistent level of geometric detail, underside of table, material, managed to capture many of overlapping geometry and really fine details.

What didn’t work: Inside of roof (more coverage needed), same level of detail could’ve likely been achieved with considerably fewer photographs.

Photocount – 298 (298 accepted)
Photo positions/angles – 6
Photo time – 15 minutes
Post processing – None, lazy
Total reconstruction time – 1.5 hours
High detail polycount – 38.2M tris

Barbeque

April 5th 2020

What worked: Overall shape, material (considerating how much shine), base and fine areas such as bolts.

What didn’t work: Lots of errors between geometry and around edges (needed considerably more closeup coverage).

Photocount – 298 (298 accepted)
Photo positions/angles – 6
Photo time – 20 minutes
Post processing – None, lazy
Total reconstruction time – 1.5 hours
High detail Polycount – 59.5M tris

Lawn Mower

April 6th 2020

What worked: Material came out great, overall shape captured well, lots of fidelity in really tricky areas (motor cage, cables, fabric).

What didn’t work: Wheels (not enough coverage), holes in fabric (poor coverage), many photos ignored due to poor focus, finer details lost in decimation (better settings required).

Photocount – 267 (225 accepted)
Photo positions/angles – 6
Photo time – 15 minutes
Post processing – None, lazy
Total reconstruction time – 2 hours
High detail polycount – 71.9M tris

Pot of plants and stuff

April 7th 2020

What worked: Built on approach used in lawn mower by focusing more strategically on key details. Twigs, leaves etc all came out really well, super high detail on some of the stones and even a couple of leaves/stems.

What didn’t work: Holes in pot (more coverage needed), stems (moved too much in wind, better to isolation), leaves (needed better focus in photographs).

Photocount -350 (323 used)
Photo positions/angles – 8
Photo time – 25 minutes
Post processing – Shadow/highlight removal, sharpening
Total reconstruction time – 2 hours
High detail polycount – 62.8M tris

Sun Dial

April 8th 2020

What worked: Again used approach from lawn mower, thus really fine geometry came out well and the metal dial, especially as well as the overall shape. Grass shaped up pretty well, too.

What didn’t work: Underside still weak (more coverage needed, struggled to get underneath), too many photos (likely an excessive amount for the detail I captured. Would’ve been wise to invest more time on weaker areas).