Pix4d Vs Agisoft - Model Recreation. Trees Missing in Agisoft.

[CanadianOne]

When creating a model using the same images Ive taken of a house and yard, the results are quite different. Pix4 shows 95% of the trees in the area while Agisoft only shows $30 and Agisoft seemed to created the house a bit better then Pix4D. Has anyone else noticed this? Is this because I did not use the right settings on Agisoft?

[JMR]

Hello Canadian
trees might have been removed by "depth filtering" which is by default set as "agressive". This filter is intended to reduce noisy points. If you lower or disable this filter you will likely have your trees back along with some other points floating in the air but sometimes others under the terrain also
Best
GEOBIT

[stihl]

Different software, different algorithms, different output.

A workaround is to generate a DEM on a "medium" quality output. Export this file and then rerun the dense cloud generation in a "low" quality output.
The word quality is a bit misleading as this refers to the scale for the input photos.
Low means that the original images get downscaled 1/8th of their original size and they are then used to find dense points.
Medium means the photos get downscaled 1/4th of their original size.

This downscaling of images means that photoscan will more easily find dense points in vegetation.
As vegetation moves because of the wind in each image, using the original image sizes will likely yield bad results because at that scale the pixels of vegetation moves too much.
If you down scale them, the vegetation becomes less visually complex and thus more easily to generate a tree.
This same effect happens if you increase your flight altitude.

After you've rerun your dense cloud I would export this low quality DEM as well. Then in a GIS like Globalmapper you can filter out everything except the trees and then merge these two DEMs into one product.

As a final note; do you think it's fair to discuss a competing photogrammery package on the Agisoft forums?
As Pix4D doesn't even have a community forum this just seems like an unfair advantage for px4d as their product downsides and problems (which it has many of) never get highlighted...

[CanadianOne]

Different software, different algorithms, different output.

A workaround is to generate a DEM on a "medium" quality output. Export this file and then rerun the dense cloud generation in a "low" quality output.
The word quality is a bit misleading as this refers to the scale for the input photos.
Low means that the original images get downscaled 1/8th of their original size and they are then used to find dense points.
Medium means the photos get downscaled 1/4th of their original size.

This downscaling of images means that photoscan will more easily find dense points in vegetation.
As vegetation moves because of the wind in each image, using the original image sizes will likely yield bad results because at that scale the pixels of vegetation moves too much.
If you down scale them, the vegetation becomes less visually complex and thus more easily to generate a tree.
This same effect happens if you increase your flight altitude.

After you've rerun your dense cloud I would export this low quality DEM as well. Then in a GIS like Globalmapper you can filter out everything except the trees and then merge these two DEMs into one product.

As a final note; do you think it's fair to discuss a competing photogrammery package on the Agisoft forums?
As Pix4D doesn't even have a community forum this just seems like an unfair advantage for px4d as their product downsides and problems (which it has many of) never get highlighted...

Interesting, thank you for the help. I should of stated that I am uploading the exported OBJ Models to Sketchfab. Also would this be the same problem when having the roof of a house of some smaller holes appearing in the Mesh and Text, but viewing the dense cloud there are no holes in the roof.