Depth Estimation on Single Camera with new Depth Anything State-of-the-art Model

Рет қаралды 14,515

Күн бұрын

Inside my school and program, I teach you my system to become an AI engineer or freelancer. Life-time access, personal help by me and I will show you exactly how I went from below average student to making $250/hr. Join the High Earner AI Career Program here 👉 www.nicolai-nielsen.com/aicareer (PRICES WILL INCREASE SOON)
You will also get access to all the technical courses inside the program, also the ones I plan to make in the future! Check out the technical courses below 👇
In this video 📝 we are going to take a look at the new state-of-the-art depth anything model on monocular images. We are going to go over the benchmarks, GitHub repository and I'll also show you how to set it up and run it in your own custom python scripts and environments.
Code: github.com/niconielsen32/dept...
If you enjoyed this video, be sure to press the 👍 button so that I know what content you guys like to see.
_____________________________________________________________
🛠️ Freelance Work: www.nicolai-nielsen.com/nncode
_____________________________________________________________
💻💰🛠️ High Earner AI Career Program: www.nicolai-nielsen.com/aicareer
⚙️ Real-world AI Technical Courses: (www.nicos-school.com)
📗 OpenCV GPU in Python: www.nicos-school.com/p/opencv...
📕 YOLOv7 Object Detection: www.nicos-school.com/p/yolov7...
📒 Transformer & Segmentation: www.nicos-school.com/p/transf...
📙 YOLOv8 Object Tracking: www.nicos-school.com/p/yolov8...
📘 Research Paper Implementation: www.nicos-school.com/p/resear...
📔 CustomGPT: www.nicos-school.com/p/custom...
_____________________________________________________________
📞 Connect with Me:
🌳 linktr.ee/nicolainielsen
🌍 My Website: www.nicolai-nielsen.com/
🤖 GitHub: github.com/niconielsen32
👉 LinkedIn: / nicolaiai
🐦 X/Twitter: / nielsencv_ai
🌆 Instagram: / nicolaihoeirup
_____________________________________________________________
🎮 My Gear (Affiliate links):
💻 Laptop: amzn.to/49LJkTW
🖥️ Desktop PC:
NVIDIA RTX 4090 24GB: amzn.to/3Uc7yAM
Intel I9-14900K: amzn.to/3W4Z5Cb
Motherboard: amzn.to/4aR6wBC
32GB RAM: amzn.to/3Jt2XVR
🖥️ Monitor: amzn.to/4aLP8hh
🖱️ Mouse: amzn.to/3W501GH
⌨️ Keyboard: amzn.to/3xUGz5b
🎙️ Microphone: amzn.to/3w1F1WK
📷 Camera: amzn.to/4b4Ryr9
_______________________________________________________________
Timestamps:
0:00 Intro
0:36 Project Page
3:01 Github
4:32 Code Walkthrough
10:24 Results on video
10:48 Results on webcam
13:45 Outro
tags:
#Depthmap #DepthEstimation #DepthAnything #Zero-Shot

Пікірлер: 57

@NicolaiAI 5 ай бұрын

Join My AI Career Program 👉 www.nicolai-nielsen.com/aicareer Enroll in My School and Technical Courses 👉 www.nicos-school.com

@Rahi404 5 ай бұрын

Hey there! Thanks for introducing me to the Marigold model. After testing it, I found it wasn't ideal for real-time applications, and even Midas fell short for my thesis project. Luckily, your videos led me to the perfect solution - "Depth Anything". I'm excited to fine-tune it for my project. Your content has been a huge inspiration for my work in Computer Vision. Keep up the great work! 👏👁‍🗨

@NicolaiAI 5 ай бұрын

Thanks a lot for the Awesome words mate! Yeah this depth anything can now run in real time and with great performance compared to Midas. Huge leap and hope you can use it for some cool stuff!

@a.........1._.2..__..._.....__ Ай бұрын

Everything im learning about my current project keeps bringing me to your videos. Thanks for posting these indepth videos. Has helped a lot during my learning process.

@NicolaiAI Ай бұрын

Thanks a ton man! Happy u find them helpful

@a.........1._.2..__..._.....__ Ай бұрын

For those who might be curious. I'm attempting to combine yolo and depth estimation to identify products on a shelf and how many are required to fill vacant locations. Got annoyed having to manually recall, so the engineering brain kicked in :)

@Chalce-dony 5 ай бұрын

I have setup both midas and Depth-Anything with your instructions and i am using it to do inference on rtsp stream. But the Depth-Anything model is being much slower than Midas model in my setup. What could be the reason of this?

@abdullahal7512 5 ай бұрын

any idea how i can use/implement this algorithm for certain case for my banchelor thesis ?

@rajmehta8170 3 ай бұрын

How can I get depth information from heat map created so that I know if particular object is how much far or near relative to camera ?

@aleven27 2 ай бұрын

is there anyway you can create a code where you can import a video and export it out as a depth map, please let me know 🙏

@unveil7762 4 ай бұрын

Awesome! How can i get the cuda memory address before to go the cpu? Is a waste of process to the image processing in cpu. Anyway awesome video!!

@eugenegamez3112 Ай бұрын

So, let's say I have a single normal camera and I take a photo of a certain object. Is it possible to get real-world coordinates of that object from the photo using this model?

@entrepreneerit4490 5 ай бұрын

Can these be converted to actual distances or do we need to use the metric depth model for that?

@teetanrobotics5363 5 ай бұрын

One of the best channels for ai, computer visiion and deep learning

@NicolaiAI 5 ай бұрын

Wow, thanks a lot for the nice words! Appreciate all of u

@nourhanahmed9757 5 ай бұрын

can we print out the depth map as we did in the MiDaS model ?

@steffen3536 5 ай бұрын

Thank you! Are these depth estimators fast enough to run on edge devices like iphones?

@NicolaiAI 5 ай бұрын

They are the fastest by far but will need to be optimized and exported to a format that’s supported by apps. But in that case it can run real time but not easy to do

@steffen3536 5 ай бұрын

Ah I see thanks ! :)

@lyafpv9056 5 ай бұрын

I wanted to ask, is there already a version for stereo vision? And how can one calculate distances with an AI detection overlay to obtain object distances? thanks for this super video

@NicolaiAI 5 ай бұрын

Thanks a lot, then you will have to use metric depth from the model, check out their GitHub repo! Might do more videos about that as well

@entrepreneerit4490 5 ай бұрын

@@NicolaiAI please do! I'd love to understand how metric depth works.

@AAA-xc4zm 5 ай бұрын

Hi Nicolai. Is there a model that can output the depth information of a point if we provide the pixel coordinates of that point?

@vipulpardeshi2868 5 ай бұрын

I also want that using Single camera

@RasitEvduzen 5 ай бұрын

How can we get scaler distance value for specific object from depth video if it's possible?

@ed6280 5 ай бұрын

same question

@ajarivas72 5 ай бұрын

@@ed6280 AI will get it with no problem.

@NicolaiAI 5 ай бұрын

You will need to extract the positions. Normally you would come it with a segmentation model to do that

@ashutoshshukla4680 5 ай бұрын

So does it gives default depth in meters can we get that without training@@NicolaiAI

@malek3764 5 ай бұрын

Thanks for this video! I just need to ask on what hardware did you run this for real-time performance? And what FPS (or inference time in milliseconds) did you get for each model?

@NicolaiAI 5 ай бұрын

Thanks for watching! Just from raw model I get around 50 fps or so. Can definitely be bumped up with optimization and deleting the visualizations etc. This is only a 25mb model so very small vision model! Can run pretty fast. I’m running this on a 3070

@malek3764 5 ай бұрын

@@NicolaiAI wow that's actually impressive! thanks alots

@NicolaiAI 5 ай бұрын

@malek3764 yeah that’s their small model they also have way larger models which can’t run real time. But now we have amazing results in real time

@malek3764 5 ай бұрын

@@NicolaiAI based on ur experiments with the models, how big is the gap in quality between the small one and the larger versions?

@NicolaiAI 5 ай бұрын

@@malek3764 not much! Definitely go with the smaller models unless you do some 2D to 3D stuff and those things where you want the highest detail

@vipulpardeshi2868 5 ай бұрын

Hey i just wanna know what will be this models performance on Embedded Devices like Jetson or Raspberry Pi

@NicolaiAI 5 ай бұрын

Can prob run a few frames on a jetson

@vipulpardeshi2868 5 ай бұрын

Thanks for the reply😊. Just one more question Can we get the depth data in like centimetres using this model

@keshav2136 5 ай бұрын

The video is nice. Why don't you provide required links in the captions?

@NicolaiAI 5 ай бұрын

Thanks! Will upload to my GitHub today! Had some urgent stuff that came up right at release

@keshav2136 5 ай бұрын

@@NicolaiAI Alright, fair enough, and I've seen your other videos as well for Stereo Cam Calibration and now this Mono Camera Depth Estimation, it would be SOO helpful if you provide links to the targetted GitHub repos or links in the description or Comment section. We literally are the ones who watch your videos full from the beginning to the end. So please, take care of that. Thanks 😃

@NicolaiAI 5 ай бұрын

I appreciate all of you! I promise ill do my very best and do that going forward @@keshav2136

@NicolaiAI 5 ай бұрын

github.com/niconielsen32/depth-anything

@cg.man_aka_kevin 4 ай бұрын

Can this code run in Python?

@NicolaiAI 4 ай бұрын

Yup this is running in python

@cg.man_aka_kevin 4 ай бұрын

@@NicolaiAI Owh, because I saw the icon software doesn't look like Python... 😅😅😅

@ruanjiayang 4 ай бұрын

What is "relative depth"?

@donrikk8546 3 ай бұрын

depth relative to the camera/viewer i would assume. since depth can be captured in different ways i think the term relative in this case is just a designation for the camera it is being captured on mimmicking depth youde produce using both of your eyes naturally, thats why this depth map is also used in 2d to 3D video conversions cuz it helps mimmick stereo view like the way studios do it

@kirkeby7875 5 ай бұрын

Siger ikke hele ideen til min bachelor lige røg i vasken, men jeg bruger stereo-vision med IR-projektør... Fedt man nu kan bare med et enkelt kamera... yay! 🤣🤣

@NicolaiAI 5 ай бұрын

Haha sorry 😂

@kirkeby7875 5 ай бұрын

@@NicolaiAI Tak for altid at poste den nyeste viden, det holder os på dupperne 💪

@NicolaiAI 5 ай бұрын

@@kirkeby7875 mange takker for at følge med!