-2

I do product photography of furnitures and my client is asking for the products to be placed in interior background like the one on amazon, as i do photoshoot in their factory and not in a real house, its difficult to create interiors with well perspective. I tried PS generative fill but its too tedious and time consuming. I do have subscription for stock 3d models but I'm not able to place the furnitures photos in them properly.

enter image description here

Duarte Farrajota Ramos
  • 59,425
  • 39
  • 130
  • 187
Mr. VOID
  • 1
  • 1
  • 1
    "I'm not able to place the furnitures photos in them properly.". Why, what's wrong? Care to elaborate? – Leander Mar 15 '24 at 07:35
  • 1
    Use FSpy to estimate camera position. – Leander Mar 15 '24 at 07:36
  • Check light interaction with 2d elements. – Leander Mar 15 '24 at 07:37
  • @Leander thanks, but as you can see in the image that i just attached, the wardrobe is a real photo shot by me and background is created in photoshop, i want to use the already created 3d interior and just place the furnitures photos in it, using just simple geometry for shadows. – Mr. VOID Mar 15 '24 at 08:46
  • Please try to be more descriptive with your issue. Do you need a photo turnaround or a 3D model? Do you need Augmented Reality integration? Do you want photo as billboard, create 3D from photo or use yor images to customize existing 3D assets?

    As you can see, the answer can be really different based on the real needs.

    Once you refine the description down to the technical problem, i suggest you to change the question to better reflect the issue and reach to more people to help you.

     – Luis Quintana Mar 15 '24 at 08:50
  • If I understand correctly you are doing the opposite. Usually, you need to insert a 3D model of furniture into a real photo of the room. And not make a whole room around a photo of his furniture. What your client is asking for doesn't make sense to me. That's not how it's done. – Shubol3D Mar 15 '24 at 09:41
  • 2
    What the client is asking for totally makes sense. – Leander Mar 15 '24 at 13:53

1 Answers1

1

If you are taking the photos yourself that's at least possible. You would have 3 main tasks/problems to solve:

  • You need to match your camera and lens settings and properties and position and orientation in relation to your subject to your 3d scene. You should adjust sensor size of Blender's cameras to match your real camera as well as lens focal length. You can find your camera model's sensor size online. Lighting will also need to match as closely as possible, luckily you can control the lighting in both cases.

  • You will have to have a way to deal with the shadows. You could attempt to take the shadows from the photographs(that might be a hard task) or model rough shapes of the furniture to cast the shadows in your 3d scene. It's possible to put the objects for casting shadows in one collection and enable Indirect Only to render only their contribution to indirect lighting and shadows in the scene while hiding objects themselves from the camera:

enter image description here

Note that this is hidden by default and to turn this option on in the View Layer mode of the Outliner you might need to enable this in the filter menu in Outliner's header:

enter image description here

  • You will have to deal with color management to match color between your camera and the render. Since you have the option of photographing in raw format, that might be a lot of help. If you used raw photographs you could convert them to EXR format in linear color space with REC.709 primaries that is the internal color space used by Blender to do rendering calculations(you could use Darktable for that(it's awesome and free)). EXR format will import to Blender as is, without any color management(that would be desired in case of linear color space) and so you can then use Blender's color management like AgX correctly when rendering. It doesn't need to be AgX and you can use whatever you want. You can use other software to cut the objects out of the background, or even do that in Blender by simply putting the images on mesh planes and cutting the mesh geometry with something like knife tool, or other ways. You would obviously have to use some sort of shadless shaders and possibly want to disable all object ray visibility except camera for the planes not to interact with the lighting of the scene:

enter image description here

Or you could put also composite the objects on top of the background in the Compositor. Linear EXR format would still make sense. It might be easier to adjust lighting of the scene with objects in the scene while previewing ths scene in the viewport rendered mode though.

Photogrammetry could help here since you do have access to the products. Figuring out a suitable workflow with photogrammetry would require a lot of initial effort and some additional work continuously, but it may be possible to somewhat automate a lot of things with it and add some accuracy to the process. It's possible to get camera position, orientation and properties from it, it's possible to get rough 3d models or at least good reference for them. It would still be quite time consuming. You could take the main shot and a 3-5 additional photos per product for this to work. You would need additional software for it like Agisoft Metashape(very expensive, but definitely works), or Meshroom(free) and import camera data and models from them to Blender.

It's also possible to use Blender's camera tracking and solving functionality in similar fashion for a sequence of photographs instead of video frames, but it requires significantly more(I mean really a whole lot more) manual work to set up markers for matching points in different photographs. That would probably not be practical in this case.

Martynas Žiemys
  • 24,274
  • 2
  • 34
  • 77