Loading video...

Video Failed to Load

Go Home

SpatialTracker Tracking Any 2D Pixels in 3D Space Recovering dense and long-range pixel motion in videos is a challenging problem. Part of the difficulty arises from the 3D-to-2D projection process, leading to occlusions and discontinuities in the 2D motion domain.

_akhaliq's profile picture

AK

460,253 subscribers

77,162 views • 2 years ago •via X (Twitter)

Science & TechnologyEducation
LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Is your game 2D or 3D? Yes and also in pixel mode
0:13

Is your game 2D or 3D? Yes and also in pixel mode

Dark Light

31,860 views • 11 months ago

CG Hitchhiker shared Part 2 of their “2D look, 3D rig” tests, pushing a stretchy cartoon style with clever rig tricks to keep that flat, 2D feel in motion 🎬 See more:
0:16

CG Hitchhiker shared Part 2 of their “2D look, 3D rig” tests, pushing a stretchy cartoon style with clever rig tricks to keep that flat, 2D feel in motion 🎬 See more:

Jettelly Inc.

36,383 views • 4 months ago

I've figured out how to constantly switch between 2D and 3D in order my gaslight the chat. I can start out as a 2D model then zoom out to reveal we were in 3D the whole time.
0:15

I've figured out how to constantly switch between 2D and 3D in order my gaslight the chat. I can start out as a 2D model then zoom out to reveal we were in 3D the whole time.

shindigs

54,090 views • 1 year ago

Visualizing bird sounds in 2D and 3D
0:16

Visualizing bird sounds in 2D and 3D

Massimo

249,116 views • 7 months ago

fromCompToSurface() makes it easy to convert 2D effects into 3D space in After Effects without all the math.
0:14

fromCompToSurface() makes it easy to convert 2D effects into 3D space in After Effects without all the math.

François Lefebvre

12,956 views • 27 days ago

Roundhouse kicks feel so weird in 2d and 3d at the same time
0:12

Roundhouse kicks feel so weird in 2d and 3d at the same time

Screenbound - A game about being distracted

24,911 views • 2 years ago

They’ve really mastered the art of mixing 3d and 2d
0:12

They’ve really mastered the art of mixing 3d and 2d

Agnis

491,211 views • 3 days ago

Made a little experiment with mixing a 3D element in a 2D panel on VisionOS. I like how such a simple thing makes the whole UI feel 3D: it reinforces the apparent fact that 2D is a subset of 3D, and all interactions span there. Created with with SwiftUI and RealityKit.
0:15

Made a little experiment with mixing a 3D element in a 2D panel on VisionOS. I like how such a simple thing makes the whole UI feel 3D: it reinforces the apparent fact that 2D is a subset of 3D, and all interactions span there. Created with with SwiftUI and RealityKit.

Oleg Frolov

13,426 views • 1 year ago

the transition from 2d to 3d cutscene was done really well
0:12

the transition from 2d to 3d cutscene was done really well

Marko

88,403 views • 3 months ago

This is excellent, Victor Verdugo! 😍 Victor V. created this #Houdini exploration to depict the "waste we leave." He used Cloth and ShapeMatch constraints from Vellum. Victor is a 3D Motion designer who either animates 2D designs or creates 3D pieces from scratch.
0:18

This is excellent, Victor Verdugo! 😍 Victor V. created this #Houdini exploration to depict the "waste we leave." He used Cloth and ShapeMatch constraints from Vellum. Victor is a 3D Motion designer who either animates 2D designs or creates 3D pieces from scratch.

Houdini.School

12,920 views • 10 months ago

3D to 2D, The camera is a beautiful liar. #3DAnime #IndieAnime #Rig
0:11

3D to 2D, The camera is a beautiful liar. #3DAnime #IndieAnime #Rig

Okawaii

129,029 views • 1 month ago

Monocular pose estimation has gotten really good Grab any 2D video and transfer the performance to a 3D character
0:11

Monocular pose estimation has gotten really good Grab any 2D video and transfer the performance to a 3D character

Bilawal Sidhu

27,074 views • 1 year ago

We’ve added 2D and 3D vol surfaces to the Dashboard. Next up: educational threads to walk you through how to use them in your trading process. Stay tuned.
0:16

We’ve added 2D and 3D vol surfaces to the Dashboard. Next up: educational threads to walk you through how to use them in your trading process. Stay tuned.

Menthor Q

40,367 views • 1 year ago

Ferrari F40 and 2D motion graphics. Composed in After Effects. #Ferrari
0:17

Ferrari F40 and 2D motion graphics. Composed in After Effects. #Ferrari

Hieu Vu Duc

128,026 views • 5 months ago

How to turn any 2d object into 3d!
0:16

How to turn any 2d object into 3d!

xarr

59,259 views • 1 year ago

This looks 2D... but it's 3D 🤯This 2D animation technique in Blender is absolutely BRILLIANT! ✨ Yana Zavialova (CG Hitchhiker) creating old-school 2D-looking animation with 3D rigged models? That's the perfect hybrid workflow! 🔥
0:10

This looks 2D... but it's 3D 🤯This 2D animation technique in Blender is absolutely BRILLIANT! ✨ Yana Zavialova (CG Hitchhiker) creating old-school 2D-looking animation with 3D rigged models? That's the perfect hybrid workflow! 🔥

3DxDEV

13,465 views • 4 months ago

2d and 3d 💜 work in progress #3DCG #HanaMiiArt #WIP
0:19

Sensitive content

2d and 3d 💜 work in progress #3DCG #HanaMiiArt #WIP

SpooX3D

63,477 views • 8 months ago

Meta releases VGGSfM Visual Geometry Grounded Deep Structure From Motion Structure-from-motion (SfM) is a long-standing problem in the computer vision community, which aims to reconstruct the camera poses and 3D structure of a scene from a set of unconstrained 2D images. Classical frameworks solve this problem in an incremental manner by detecting and matching keypoints, registering images, triangulating 3D points, and conducting bundle adjustment. Recent research efforts have predominantly revolved around harnessing the power of deep learning techniques to enhance specific elements (e.g., keypoint matching), but are still based on the original, non-differentiable pipeline. Instead, we propose a new deep SfM pipeline VGGSfM, where each component is fully differentiable and thus can be trained in an end-to-end manner. To this end, we introduce new mechanisms and simplifications. First, we build on recent advances in deep 2D point tracking to extract reliable pixel-accurate tracks, which eliminates the need for chaining pairwise matches. Furthermore, we recover all cameras simultaneously based on the image and track features instead of gradually registering cameras. Finally, we optimise the cameras and triangulate 3D points via a differentiable bundle adjustment layer. We attain state-of-the-art performance on three popular datasets, CO3D, IMC Phototourism, and ETH3D.
0:13

Meta releases VGGSfM Visual Geometry Grounded Deep Structure From Motion Structure-from-motion (SfM) is a long-standing problem in the computer vision community, which aims to reconstruct the camera poses and 3D structure of a scene from a set of unconstrained 2D images. Classical frameworks solve this problem in an incremental manner by detecting and matching keypoints, registering images, triangulating 3D points, and conducting bundle adjustment. Recent research efforts have predominantly revolved around harnessing the power of deep learning techniques to enhance specific elements (e.g., keypoint matching), but are still based on the original, non-differentiable pipeline. Instead, we propose a new deep SfM pipeline VGGSfM, where each component is fully differentiable and thus can be trained in an end-to-end manner. To this end, we introduce new mechanisms and simplifications. First, we build on recent advances in deep 2D point tracking to extract reliable pixel-accurate tracks, which eliminates the need for chaining pairwise matches. Furthermore, we recover all cameras simultaneously based on the image and track features instead of gradually registering cameras. Finally, we optimise the cameras and triangulate 3D points via a differentiable bundle adjustment layer. We attain state-of-the-art performance on three popular datasets, CO3D, IMC Phototourism, and ETH3D.

AK

96,527 views • 1 year ago

This is 3D animation powered by AI. No suits or markers needed. This 2D video was captured with GoPro cameras and converted to 3D motion data using Move Multi-Cam. #MadeWithMove
0:15

This is 3D animation powered by AI. No suits or markers needed. This 2D video was captured with GoPro cameras and converted to 3D motion data using Move Multi-Cam. #MadeWithMove

Move AI

182,792 views • 2 years ago