Learn Computer Vision using AI: kevinwoodrobotics.com/product/computer-vision-using-ai/ Code and Doc: kevinwoodrobotics.com/product/depth-anything-v2-monocular-depth-estimation-explanation-and-real-time-demo/
Excuse me, streamer. I'd like to ask how I can convert the output of the Depth Anything model into an actual depth map to obtain the real-world 3D coordinates of a specific pixel in the image. This is crucial for determining whether the model can be applied in real-world engineering projects.
Excuse me. I downloaded the file, unzipped it, and placed it in .nuke as you explained. But Cattery doesn't appear in Nuke. I don't know what's wrong! My nuke is 13.2v9.
Hi is this running slow because it's processed on a cpu ? What kind of FPS can one expect to get with a good GPU (390 24GB) . Would you recommend another depth model for real time needs ? Thanks
It’s going to depend on which model you’re using and what image size you want. Best way is to do a quick test and see. And yes I was using cpu here so that slowed it down
Thank you for the presentation. I am wondering if the DAV2 is capable of generating an accurate depth map - meaning that for example if we put a small cube in front of the camera (lets say in 20cm distance), can this model estimate the 20cm accurately? Or even +- few cm off...
@@kevinwoodrobotics am sorry i am quite novice to this could you explain this further in a video or what do i have to learn to do it i wanna estimate the distance between a car on a road and other cars or obstacles
Excuse me, streamer. I'd like to ask how I can convert the output of the Depth Anything model into an actual depth map to obtain the real-world 3D coordinates of a specific pixel in the image. This is crucial for determining whether the model can be applied in real-world engineering projects.
Learn Computer Vision using AI: kevinwoodrobotics.com/product/computer-vision-using-ai/
Code and Doc: kevinwoodrobotics.com/product/depth-anything-v2-monocular-depth-estimation-explanation-and-real-time-demo/
10:17 the ferris wheel bro😂 thanks for the video
Yess brain fart 😆
Excuse me, streamer. I'd like to ask how I can convert the output of the Depth Anything model into an actual depth map to obtain the real-world 3D coordinates of a specific pixel in the image. This is crucial for determining whether the model can be applied in real-world engineering projects.
ive been trying to "install it" as a tutorial said, just so i can use it for "face 3d" on After Effects for the past 2 hours, can anyone give ay tips?
Make sure you use Python environments and install PyTorch from official website
Excuse me. I downloaded the file, unzipped it, and placed it in .nuke as you explained. But Cattery doesn't appear in Nuke. I don't know what's wrong! My nuke is 13.2v9.
I’m not following
Hi is this running slow because it's processed on a cpu ? What kind of FPS can one expect to get with a good GPU (390 24GB) . Would you recommend another depth model for real time needs ?
Thanks
It’s going to depend on which model you’re using and what image size you want. Best way is to do a quick test and see. And yes I was using cpu here so that slowed it down
Thank you for the presentation. I am wondering if the DAV2 is capable of generating an accurate depth map - meaning that for example if we put a small cube in front of the camera (lets say in 20cm distance), can this model estimate the 20cm accurately? Or even +- few cm off...
I haven’t tried but that would be an interesting experiment
That's great! Is it possible to have the training code for the V2 model, as described in the paper? thanks.
Thanks! I haven’t seen it up to now
Hey Kevin, I bought your Code, can you run this on google collab? Having issues setting it up
Hey! I recommend running it on vs code locally if you could try that out
@@maxwellshoroye9247 you can follow my yolo v9 video where I use Google collab to run code. Those instructions should also apply
can it estimate the distance?
You can use the depth map with camera parameters to obtain distance
@@kevinwoodrobotics neat! did you test how accurate it is?
@@kevinwoodrobotics am sorry i am quite novice to this could you explain this further in a video or what do i have to learn to do it
i wanna estimate the distance between a car on a road and other cars or obstacles
@@abdelrahmanelseht7305 Im also looking at this same problem, did you find a solution?
Have you tired to feed ROS with this data?
No I haven’t tried yet but would be interesting to integrate
what is your gpu
1650
Excuse me, streamer. I'd like to ask how I can convert the output of the Depth Anything model into an actual depth map to obtain the real-world 3D coordinates of a specific pixel in the image. This is crucial for determining whether the model can be applied in real-world engineering projects.
Need to calibrate camera and use projection to get 3D coordinates