I recently upgraded from RealityCapture 1.4.x to 1.5.1 and have encountered an issue with the automated 3D model generation process. Previously, images that did not fit into the main dataset (those outside the intended circular capture area) were automatically ignored as long as the rest of the images aligned correctly. However, after updating to 1.5.1, I have noticed that these unrelated images are now being included in the reconstruction, causing issues with my models.
I am using an automated pipeline for model generation, and while it worked correctly with previous versions, this problem has become quite common for me since the update. Could you please confirm if there were any changes in the way RealityCapture handles image selection and alignment in version 1.5.1? Additionally, are there any settings or workflows I should adjust to ensure that only the correct images are used in my models?
Any guidance on how to fix this issue would be greatly appreciated. Thank you for your support!
Hello,
as you are mentioning the automated process, do you mean your script or Start button in RealityCapture?
Also, what do you mean by this: images that did not fit into the main dataset (those outside the intended circular capture area) were automatically ignored as long as the rest of the images aligned correctly?
Were or weren’t those images aligned? All aligned images are used for meshing until it is set differently. If you have unrelated images, you shouldn’t use them in the project.
Thank you for your response! I am using a script to automate the model generation. When I mentioned images “outside the intended circular capture area,” I meant those taken from different angles or closer to object than the majority of the images, but still aligned with the dataset. These images were not part of the central area but were still included in the process, causing deformations in resulting model. In the past, these types of images have not been affecting the result, and I can assume were not used in further steps after aligning, but since updating to version 1.5, it happens quite frequently and negatively affect the quality of resulting model.
I would appreciate any advice or suggestions on how to ensure only the relevant images are included in the model generation.
There is no way there weren’t used if they were aligned and they were enabled for meshing.
The easiest way is not use such images in your process (filter the images before processing them).
If you don’t know the image’s names then it is not possible to disable them for processing automatically.