3D Python Tutorial: Building 3D Models from 2D Images (Photo or AI) with 5 Libraries

[@birucao3132]

Thank you so much for the tutorial! It's super clear and helpful!

[@FlorentPoux]

You're very welcome!

[@TheSaturnized]

I would like to thank you for your wonderful tutorial. I have to admit, it was a lifesaver for my career. However, I recommend using a better depth estimation model, which is easy to find online. I replaced your current depth estimation model with an improved one, and the results were fascinating. It would be fantastic if you could create a tutorial on how to generate a coherent mesh from several point clouds. Specifically, I aim to enhance the output by creating point clouds from several angles and finally merging them together. Thank you, and I wish you the best of luck.

[@sampreethkabadi3701]

Hey! Could you help me out with what depth estimation model you used for improving it?

[@FlorentPoux]

I'm truly delighted my tutorial made a difference in your career. It fuels my passion to hear such positive feedback. 🎉 Regarding your suggestion about the depth estimation model, I'm always looking for ways to improve. I appreciate you sharing your findings and will definitely explore better models. I firmly believe in continuous learning and refinement. This is at the core of my teaching. Creating coherent meshes from multiple point clouds is an excellent goal. here is a tutorial on it, that i hope aids you: kzbin.info/www/bejne/mIakh3aMj9hkrbMsi=Scd5KkaXL0hPAjbg

[@TheSaturnized]

@FlorentPoux Dear Dr. Poux, Thank you for your insightful comment. I would be more than happy to help. Unfortunately, it seems that KZbin doesn’t allow external links in the comment section, as each time I tried to reply to your comment, it was immediately removed without explanation. As a result, my only option is to send the link to the model via LinkedIn direct message. Best regards.

[@RudraPanchal-gz8zw]

@@TheSaturnized hello brother can u share me model which u use for better accuracy and also your github code link

[@Petershd138]

I love your crystal-clear tutorials.

[@FlorentPoux]

Thanks a lot 🤗

[@RiccardoEmanuelMarinelli]

So clever!

[@FlorentPoux]

So many creative ways it can be used!

[@ShadowD2C]

as always the video is wonderful and straight to the point, I always get what I need about 3D from your channel, I have a question tho, I want to measure clearance (cm distance ) around my vertical object accurately, do you have any input for this sort of problem? do I need the actual meshs or do i just stop with the pointcloud?

[@mohammadqalaji8676]

Amazing Tutorial, thanks :D

[@FlorentPoux]

Glad ou like it! Thanks for the kind words!

[@elyntro]

Hello , Thanks for the video, but the image generated from sample 3d computer image is noisy so how to clean it make look like exactly realistic ? is there any library to make this happen ?

[@FlorentPoux]

Hey, with pleasure! happy it helps! To clean it, you need to go through some noise filtering, like using statistical distance to neighbor points. I will make another video on the topic then.

[@jamalkhan5431]

Can i estimate road layer thickness using this?

[@danielfavero79]

Excellent tutorial.

[@FlorentPoux]

Glad you liked it!

[@kurisutinahiyajosafinaillu6595]

Thank you for your wonderful work. But there seem to be a problem i am facing when following your tutorial on colab environment. In the "3D Point Cloud to Mesh" step, when i convert my point cloud to a mesh model it lose all of it color, leaving me with a grey mesh model that doesn't have any color. How did you managed to create a mesh model with the original color from the image

[@muhammadshareef780]

hey bro can you share me your code i did the same but its not creating the cloud points. Ill be grateful if you share it. Thank you

[@FlorentPoux]

thanks! hmmm, weird, but essentially, you use the source rgb image and project the color on top for that.

[@kurisutinahiyajosafinaillu6595]

@@FlorentPoux I think the problem is that I am following your tutorial on Colab instead of running it locally. In your tutorial, when you visualize your point cloud and mesh, you used Open3D's draw_geometries method, but unfortunately, this method doesn't support the Colab environment. So, I had to use an alternative, which is draw_plotly. However, I believe draw_plotly only supports color for point clouds, not meshes. I'm still trying to find a way to solve this issue, but I haven't had any luck so far.

[@chandrutimepass5321]

what are the minimum system requirements

[@najmussaqib7542]

What software are you using for the work flow diagram

[@FlorentPoux]

I usually use lucidcharts

[@godislove3339]

thank you!

[@FlorentPoux]

You're welcome!

[@beratbaspinar]

Can we export the generated 3D model in STL format?

[@FlorentPoux]

Yep you can, do you want to print some cool stuff :)?

[@aldenpedulla5853]

Following along with all your code but I'm getting "The kernel died, restarting..." on step 10. Any suggestions? Thanks!

[@prabhatmanna24]

Facing the same Problem

[@FlorentPoux]

Hey, this is a memory limitation issue (likely), maybe try another environment, such as through google colab to see if this solve the issue?

[@prateeksrivastava9123]

Hi Florent, Can I get the Code so that I can use it and see the results in my system?

[@alanjose5706]

Sir, my jupyter notebook is crashing while creating o3d point cloud

[@FlorentPoux]

Indeed, I think switching to a non-server IDE would help you iterating

[@ZahraNabizadehShahre-Babak]

Thanks for your useful video. Does the .obj generated file have the texture?

[@FlorentPoux]

Thanks! Yes it does have a texture indeed

[@Abhi9av-nk8fp]

Where can i get the preprocessing code?

[@FlorentPoux]

you can get it from the academy (learngeodata.eu)

[@juanfranciscovarvello6836]

Hi Florent! I'm launching a startup that offers 3D menus for restaurants, cafes, and more. Currently, I'm researching how to create 3D models from images of dishes taken from various angles. Could you recommend any tools for this process? It's essential to maintain accurate colors and ensure high-quality models so that clients can navigate and examine details with zoom capabilities, etc. Thank you for your help, and I appreciate the valuable content you provide!

[@FlorentPoux]

Sounds great! So, I think you could turn yourself towards turntables and automate the scanning process with a lightbox to control lighting. then, you can use 3D Gaussian Splatting for the visual quality of the rendering you would get from it.

[@rajanavenkatmohit1599]

Thaks for the video, I want to make a 3d model with multiple 2d images from different views with stereo and feature matching. Could you please guide me for that

[@FlorentPoux]

I added that to the roadmap of todo videos 🌞

[@Shira-Tora]

when running the cell 10 creating o3d point cloud, the console said the kernel died, restarting kernel. anyone can help?

[@FlorentPoux]

this means you have an instability with open3d, the best case is to clean your environment and recreate one from scratch

[@MohammadKazemi-yd1jl]

where can I access the code?

[@FlorentPoux]

in the comment there is a drive folder with code

[@manojjonwal9640]

​@@FlorentPouxj

[@shiqiliu1989]

@@FlorentPoux Thank you for the clear tutorial! however, I didn't find the drive folder either..

[@kriptonian0]

@@FlorentPoux and where is that comment ?

[@adityaranjan5078]

Sir , how we convert 3d images to floor plan

[@HarryStylesFanSansy]

lmao got the same internship assignment???

[@adityaranjan5078]

@@HarryStylesFanSansy Have you any idea about this?

[@FlorentPoux]

A Very good idea indeed! And this is a vast topic, I will heavily dive on it next month, but basically, decompose the process in (1) classification, and (2) vectorization (regression)

[@malduffin8310]

For anyone trying this on macOS on an M2 ( possibly other Mx chipsets ), the output will just be a black image. The exact same code works correctly on Windows.

[@guliverpotsangbam3292]

so what is the solution for this? I am also working on macOS silicon M2

[@jainamdoshi7109]

Did anyone find access to code ?

[@MrAlipatik]

not me

[@Abhi9av-nk8fp]

@@MrAlipatikdid you get the code?

[@MrAlipatik]

@@Abhi9av-nk8fp i got it. Just use copilot or chatgpt and screenshot the code and convert it to text