Title:نسل ۳۶۰ AI برای طراحی های شهری در VR

۳۶۰ و VR AI برای طراحی های شهری به شما نشان خواهم داد که چگونه صحنه های شهری یا داخلی را بگیرید و آنها را با استفاده از هوش مصنوعی به طرح های جدید ۳۶۰ تبدیل کنید. استفاده از شکل ساختمان در عکس موجود و هدایت هوش مصنوعی انتشار پایدار با طراحی باعث می شود که طیف گسترده ای از سبک ها به سرعت آزمایش شوند. ۳۶۰ ها برای مدت طولانی جایگزینی سریع و موثر برای ساخت صحنه های سه بعدی بوده اند، زیرا می توانند بدون مدل یا تنظیمات فنی، تجربه ای فراگیر را ارائه دهند. مزیت اضافه شده این است که می‌توان آن‌ها را هم در یک وب‌بروور و هم با هر هدست VR استاندارد به راحتی مشاهده کرد. برای این گردش کار، یک تصویر پایه ۳۶۰ از فضای شما مورد نیاز است. اینها را می‌توانید با دستگاه‌هایی مانند دوربین ۳۶۰ در هر محیط شهری یا داخلی که می‌خواهید تغییر دهید، بگیرید. راه اندازی این دستگاه های ۳۶۰ دستی مانند ریکو تتا آسان است و می توان آنها را از راه دور فقط از طریق تلفن هوشمند خود فعال کرد. دستگاه ریکو تتا ۳۶۰:
https://theta360.com/uk/

هدست واقعیت مجازی Oculus Quest 2: https://www.meta.com/gb/quest/products/quest-2/

اکانت اینستاگرام: https://www.instagram.com/urbandecoders/

نصب و معرفی هوش مصنوعی انتشار پایدار: https://www.youtube.com/watch?v=14vRdWQAco0&t=21s

۰۰:۰۰ Intro 00:38 360 Cameras 01:01 360 Images 01:16 Stable Diffusion intro 01:47 نصب برنامه های افزودنی ۰۳:۳۵ تولید تصویر AI 360 06:11 ویژگی Inpainting 06:38 06:38 ارتقاء داده ها ۳۰۶۰ من ۳۰۶۰ افزودن تصاویر :۵۸ افزودن تصاویر به Quest 2 08:13 اجرای ۳۶۰ در VR 09:20 خلاصه #ai #معماری #طراحی مولد #اسکیس #طراحی داخلی #controlnet #360 #vr #quest2

قسمتي از متن فيلم: I’ll show you how to take Urban scenes and transform them into new 360 designs using AI utilizing the shape of a building from an existing photo and guiding is using stable diffusion AI a large variety of styles can be quickly tested 360s have been for a long time a quick

And effective alternative to building 3D seats as they can provide immersive experiences without a model or tactical setup added benefit is that they can be viewed both in a web browser and of any standard VI headset for this workflow a base 360 image of your space is needed this can be taken

Yourself with devices such as a 360 camera in any Urban on tier environment you would like to transform these handheld 360 devices such as the recover better are easy to set up and can be remotely activated through just your smartphone for this example I’ll use a stock 360

Image that you can find plenty online I’ll select this dilapidated Urban C with some run-down buildings which I’ve seen better days and see how we can rejuvenate this space next step is to run the AI algorithm we’ll be using the stable diffusion with control net extensions which I’ve shown

How to use install in a previous video so please take a look at that first essentially this video explained that there are two main ways to run stable diffusion one is on a cloud with a paid service such as one diffusion and the second is install and run

Locally using GitHub and then clicking the web URL to run is low key on your browser with this up and running there are two essential extensions needed which might not be already installed but it’s very easy to do first is Koya additional Networks for this go over to the extension tab click

On the available Tab and scroll down to find The Courier additional networks and hit install tab you need to press the apply and restart UI for this to refresh in exactly the same way I’ve been controlling it on the extensions select this option and hit install download two train models from the net

To be used to this the first is to go to civitai.com and download the latest Labs 360 model which is trained on 360 images as expected this needs to be pasted into your stable diffusion folder if running locally under models and in the law of order the second model is the realistic Vision

Trained AI model which again as the name suggests is trained to give more photorealistic results this is optional but I found it provided good results for architecture imagery or this go to hugging face website and download this checkpoint and go back to your stable diffusion folder under models and under stable

Diffusion paste this file here now you’re ready to get going with the image generation and your main stable diffusion AI head over to the right side of the screen under the generate button there’s a tiny icon in the middle click this button and then you can go over to the lower tab

And you see the latent Labs 360 model here click on this and text will appear at the start of your prompt to enable 360 generation you can also load the realistic Vision 2 checkpoint model from the top here if you like or keep the default 1.5 checkpoint

Hit that same tiny button again on the right to return to the main UI and it is time to add the base image click the control net tab under the additional Network section here you can drag and drop your base 360 image press the enable button and add a depth

Option to both the model and the preprocessor this essentially creates a depth map on which the image acts as a guide to help the AI keep a similar space informed if you wish for the AI to follow the images less you can move the control weight slider to a lower value

Selecting the tile check box is also very important as it creates the seamless images with this resize model can be set to just resize and the width and the height of the image should be a two to one proportion if you have a good GPU set the width to

۲۰۴۸ and height to 1024 the final step involves adding both positive and negative problems similar to my previous video add a description of what you would like to see along with the style and quality so here I would like to try a modern contemporary concrete residential building and just hit the generate button

You see that it follows the form of the building very well with modern looking windows and concrete texture however the foreground has a strange mix of grass and concrete as always AI generation is an iterative process so keep amending the prompts and adding to the negative parts to remove unwanted elements

I’ll fast forward this process but you can see that I’m changing the materials and the foreground and adding some architectural Styles such as chipperfield zahadid and kangakuma also feel free to change the sampling methods to get different results and change the sliders for the CFG scale and control weight

The lower the value to the CFG scale the more freedom the AI has to invent and the lower the value on the control weight the less the AI sticks to original form of the input image so these help very dissolves greatly As you can see the results are far from perfect so there’s always a need for manual amendments at this point one way is to selectively change areas using an in-paint feature once you’ve generated an image you like you can press the send in paint button underneath the image which then passes

This image to the inpane tab here you can brush over the areas which you would like to edit I’ll just change the pump to what you’d like to change okay for example I want to add a Japanese garden and then hit generate again this way you can iterate over just the

Selected Parts as many times as you like the next few sections are additional workplace to upscale your image and view them within your VR headset within the UI under the extras tab there’s a useful option to patch upscale or your generated images set the resize scale to 4 so the order

Images will become 8K resolution and select the upscaler model and enter the current size of the images which is 2048 by 1024 if you click on platform directory option and paste in the file path to your folder the generated images and then an output path you can resize everything all at once

Hit generate and they will always save out automatically in this final section I will show you how to view the image in VR for this you need to add some metadata to the images so that the vi headset will recognize automatically that these are 360 images but let’s go over to

Exif excel.com and download this app to inject the metadata into our 360s once downloaded and run it’s a simple method dragging and dropping the images into the app and hitting the add metadata button are we using the Oculus Quest 2 headset for this VR experience although you can

Do this with any headset just plug the Oculus into the PC with a cable and within the downloads folder in The Quest 2 paste all our generated 360s after that put on your headset and click the TV button on the bottom bar of your home screen on the central UI screen there’s a

Middle button which says your media press this and you can find all the 360s you uploaded since we injected the 360 metadata into the images Oculus will automatically detect these as a 360 experience so all you need to do is click and load here you can see our dilapidated base

Image and if we move to the AI generated images you can see how these buildings transform this way you can move between all your 360 images compare the various Styles while next to each other these images were produced very quickly and with minor editing so you can see

Average for example in the texturing or some of the corners in the buildings however with a bit more iteration and Photoshop edits you can see how useful this workflow is to create proposals for rejuvenated existing spaces you can see the distinctive changes between options and could be a useful

Workload The Smiths or pick styles that immediately grab your attention the example here has been of an outdoor space however works just as well for interior space which could be used for renovation the process of taking a real space and generating potential options to be viewed immersively is one that has great

Potential in the future shot actually in the near future this AI generation will happen within XR environment on the flight provides quick Generations within an interactive environment to drive decision making are we making more videos about using Ai and Technology design so be sure to take a look thank you

ID: TR8cjC0W2Nw
Time: 1684256235
Date: 2023-05-16 21:27:15
Duration: 00:09:55

منبع