Need help asap.

I have a dataset of about 2000 png images that I need to put together to make a 3d model with. My goal is to be able to convert this model to a mesh, so I can use it in blender. I have been trying to us ImageVis3d, but it won’t let me export it as a mesh. It puts the images together, and I can export them in different dataset formats, but not as mesh. Is there a program that could help me? I am lost and this needs to be done in two days.

Hello ProfUnicorn,
there are different solutions to your problem. Probably the easiest and “laziest” way would be to upload the images to 123D Catch (Autodesk); they use huge clusters to compute 3D models from images.

The problem here is 1.) the high amount of images 2.) the images get resized to 3 megapixels and 3.) the resulting amount of polygons will be around 1-1.5 million, which depending on the size of your scene and the amount of your images could be too little to sufficentely represent the object of interest.
The big advantage here is that you receive a fully textured mesh without much work on your behalf.

Another solution would be to use VisualSFM. There is a fairly comprehensive tutorial on how to use this software; large parts of the computation is done using GPU (CUDA only, not OpenCL) so it’s fairly fast. Though I don’t know if the amount of images / the resolution of the images will fit in your VRAM. Please note, you would need a Nvidia GPU to use this feature.
Good news though incase the images are too large or you don’t have the according GPU; you can deactivate the use of GPU, so it will just use your CPU.

The disadvantage here is that you just receive a colored pointcloud; further processing will be necessary.

So you will have to use some software for meshing the pointcloud; my usual recommendation for this kind of tasks is Meshlab. Using the poisson reconstruction yields usually good results, though it largely depends on the proper parameters. Filters - Surface Reconstruction: Poisson. Open the filter, under help you’ll find a short description on what the settings do. Feel free to ask more about this, when you got to this stage.

Additionally Meshlab offers the possibility to transfer the color information of the pointcloud to the mesh, it won’t look as nice as when projecting the full resolution of the images on the mesh but it’s a start.

Meshlab is capable of importing / exporting a broad range of different 3D file formats, maybe should check if it offers something that ImageVis3D can export.

Hope this helps.

p.s. choosing an appropriate title for your thread makes it more easy for people with knowledge about this subject to find it.

I just red another thread of yours, are those images a simple outline (contour) which, when stacked, result in a 3D object? If so, the above mentioned methods won’t work since they apply to normal photographic images.

Please, in future, try be more specific when describing your problem.

Thanks for your reply! And I’m not really sure what you mean by if the images are a simple outline. They are actually images for the Visible Human Project, so they are pictures of the human body. They used a machine to take slices off the body, small amounts at a time, and took picture after each slice. So I have very detailed images of this process that I’m trying to put into a 3d environment. I will try the method you suggested.

I think I now know what you ment by a contour outline. I apologize for the lack of description. As you surely know the solutions in your first post will not work for the data I have to work with, but I do thank you for showing me these programs. I will most likely use them in the future! Do you have any suggestions for my original question, though, taking into account that the images are a contour outline?

A handful of people already successfully used Blender for visualizing medical data derived from CT scans. I’ve never worked with this type of data, so i’m not much of a help here.

This might be something for you:

Otherwise just do a search on “blender medical data” and quite alot of results show up-