3d carving of a photograph

[Adamwoodward]

Hello, would anyone happen to know what is the best way to 3d carve an image from a photograph (i.e. portrait, album cover e.t.c.)?
I’m planning to do these on to 18mm thick oak at roughly 400mm square but just would like some advice on tool settings and also cutter size choice which would be best to achieve these.
I’m not looking for a particularly deep cut, say around 3mm, just enough to get the effect.
Thankyou in advance.

[WillAdams (William Adams (Carbide 3D))]

There are a couple of possible approaches:

• eschew 3D and simply cut a halftone:

Halftone generator on Cutrocket

This was announced/demoed in a limited way a while back — had occasion to reference it in a support ticket recently, so here’s the text from that: We have a halftone generator which will accept black and white images, so you will need to convert any image into a black and white picture. That is the only type of picture the generator will accept. Help file: Halftone Generator - Help

• eschew 3D, re-draw, and cut as a V carving

• re-draw, then use the drawn geometry as the basis for a 3D model

• manually sculpt a 3D model using the the photograph as a guide

• use a Large Language Model (what folks are passing off as “A.I.”) to create a height/depth map or 3D model — while not a photo, here was a recent discussion of doing this sort of thing from a pixel image:

File conversion help

I have a photo of a clock. How do I convert the image to an SVG…? and then convert the SVG to make it 3D. I want to convert from a photo… to an SVG then to carve a clock face in 3D.

It might be that there are specialty models for this sort of thing, but that’s not something which I have looked into.

[Jeffish (Jeff de Ruiter)]

You can try this image to relief app that was posted by a forum member a while back. A few of us have had success with it

Image To Relief - a Hugging Face Space by blair-johnson

[xenith (Blair)]

I actually just pushed some updates to it today. There’s a new curve tool, some performance improvements, better previews, and a bunch of bug fixes. Hope you all like it

[jtclose (Jim Close)]

I was testing the new code for a carve of a golden doodle. Previously I could not get the face to “pop”. Per instructions I completely removed the background to get the maximum depth of the dog. Just starting the process to carve it now.

[xenith (Blair)]

The curve tool should let you avoid needing to do any background subtraction. You can use it to clip the deep background and concentrate all the dynamic range on your subject.

For example, this curve clips the background and adds some additional contrast to the dog’s body without needing to do any background subtraction before uploading your image.

The base depth map has a large depth gradient fading off into the background of the image, so the dog itself looks flat in comparison.

After applying the curve, the background is clipped and all the range is spread between the deepest and shallowest parts of the dog itself.

This lets us get a more natural looking and convex head/snout region without doing any image manipulation.

Original, no background subtraction:

[Jeffish (Jeff de Ruiter)]

I really like the addition of the curve tool. I’ve been paying a bit with it on an elephant image that I’ve been trying to perfect for 3d carve and it really helps to define the parts that I want to stand out. There’s still a couple areas of the image that I am trying to figure out but I’ll work on it later and see what I can do

[jtclose (Jim Close)]

Here it is, bare pine wood. I will shellac and have the owner finish it.

[Jeffish (Jeff de Ruiter)]

@jtclose Looks awesome Jim! How thick is the pine? The depth of the head and left side really looks good.

[jtclose (Jim Close)]

1.5" thick.

I had my son in-law rip 3.5 thick boards in half.
They clean up around 1.6 unless it had a lot of character.

I spend more brain space trying to clean up stock than carving.

[Adamwoodward]

Thankyou, could you tell me how do i get the stl file to be able to be put into carbide create?

[flynnie (Ed Flynn)]

On the model tab, select the icon that looks like the ball to import an stl file.

[Adamwoodward]

Thankyou, i seem to be having trouble saving the stl file from image to 3d relief generator, i just get the option to 3d print?

[flynnie (Ed Flynn)]

Scroll to the bottom and click the blue highlighted link.

Due to file sizes it might take some time.

[Adamwoodward]

Thankyou, i will give it a go.

[Randy (Randy Gordon-Gilmore)]

Great program, Blair! Thank you for making it! I am trying to create a bas-relief plaque that never existed other than as a flat concept drawing in the early 1930’s. On the Image To Relief website, the input image is limited to 1.5 megapixels but that could be changed by “local” installations. Is there a downloadable version of your app that I could use for larger images?

[xenith (Blair)]

It is possible to run the app locally, but it requires a little bit of know-how or a willingness to learn. This isn’t packaged up nicely for an end-user. Huggingface spaces are just git repositories, and clicking the three dots in the top right will give you options for cloning the code to your local machine or example commands for running it with docker.

You could also download app.py and requirements.txt directly (with the little download buttons) and use those to set up a python environment and run the app. The README has instructions for changing the max resolution.

Minimal instructions for installing:

1. Install python from Download Python | Python.org. Make sure to select the “Add Python to PATH” option in the installer.
2. Copy the project files to a directory on your computer. If you’re downloading files directly, then you just need app.py and requirements.txt.
3. Open a terminal (“command prompt” on windows, or “terminal” on mac), and navigate to the directory where you saved the project files.
4. Create a python virtual environment. Run python -m venv venv. If you see an error about python not being found, try python3 instead of python.
5. Activate the virtual environment. On windows run venv\Scripts\activate. On mac run source venv/bin/activate.
6. Install project dependencies into your venv. Run pip install -r requirements.txt. You should see a bunch of stuff flash by on the screen as packages that the app needs are installed. Also run pip install huggingface_hub.
7. Go make an account on huggingface.co. Log-in, click on your profile photo in the top right, then go down the list and click on Access Tokens. This will take you to a page with a big “+ Create New Token” button. Select “read” for token type and then give it a name (like “my laptop”) and hit the “Create Token” button. It will display a long string of random text. This is your huggingface token, copy this into a text file on your computer and save it somewhere you’ll remember. You’ll use this token to download the weights for the computer vision model that predicts the base depth-map in the app.
8. Download the depth estimation model. From a terminal with your venv activated, run hf download "depth-anything/Depth-Anything-V2-Large-hf" --token {paste your token here}. You should see some progress bars as the model gets downloaded.

Minimal instructions for launching the app:

Once you’ve installed the dependencies, you can launch the app like this in the future.

1. Open a terminal and navigate to the directory with your venv and app.py.
2. Activate the venv, see step 5 above. You should see (venv) or whatever you named it appear in the terminal.
3. Launch the app. Run python app.py --max-resolution-pixels 4000000 (for a 4MP max resolution). You should see a url displayed in the terminal. By default, this should be http://0.0.0.0/7860
4. Open a browser and visit the url, you should see the app.

Hopefully this is helpful!

[Jeffish (Jeff de Ruiter)]

@Adamwoodward Sorry Adam I wasnt online all day. Ed’s answer is right and it usually doesnt take more than a few seconds to download… at least on my laptop

[Randy (Randy Gordon-Gilmore)]

Wow, thank you for the extensive answer, Blair.

@xenith

a little bit of know-how

That pretty well sums up my Python status. Although I use Python on a Raspberry Pi at work and a PC at home, it is really basic programming with the Python interpreter, using the GPIO to run stepper motors and optosensors. I will review your guide and see if it is something I’ll want to attempt. thanks again!

[Adamwoodward]

Thankyou Jeff, i have managed to download it, just trying now to figure out the toolpaths in carbide to get it machined, trying to bring down the cutting time a bit!