Zhen Wang

PhD student @UCLA_VMG @UCLA | Prev Intern @Meta @Google | @zuckermanbrain @Columbia.

[Zhen Wang @zhenwangwz]

@ak @eccvconf We also have a depth fusion module at each denoising step of the diffusion model to enforce geometry consistency, even at test time!

Unfortunately, none of the authors was able to present the paper in person, but DM/email is open for any questions / discussions!

[Zhen Wang @zhenwangwz]

Thanks @AK for tweeting our project - it has been accepted @eccvconf '24.
We present a multi-view depth 3d representation and show that with diffusion model, it can be used for various 3d tasks such as shape generation, completion and regularization.

[AK @_akhaliq]

MVDD: Multi-View Depth Diffusion Models

paper page: https://t.co/8tdciolVmd

Denoising diffusion models have demonstrated outstanding results in 2D image generation, yet it remains a challenge to replicate its success in 3D shape generation. In this paper, we propose leveraging multi-view depth, which represents complex 3D shapes in a 2D data format that is easy to denoise. We pair this representation with a diffusion model, MVDD, that is capable of generating high-quality dense point clouds with 20K+ points with fine-grained details. To enforce 3D consistency in multi-view depth, we introduce an epipolar line segment attention that conditions the denoising step for a view on its neighboring views. Additionally, a depth fusion module is incorporated into diffusion steps to further ensure the alignment of depth maps. When augmented with surface reconstruction, MVDD can also produce high-quality 3D meshes. Furthermore, MVDD stands out in other tasks such as depth completion, and can serve as a 3D prior, significantly boosting many downstream tasks, such as GAN inversion. State-of-the-art results from extensive experiments demonstrate MVDD's excellent ability in 3D shape generation, depth completion, and its potential as a 3D prior for downstream tasks.

[Zhen Wang @zhenwangwz]

@anne_churchland @UCLAHealth @dgsomucla Best wishes for a speedy recovery!

[Zhen Wang @zhenwangwz]

(5/n) Come by our poster on 📅Tuesday morning (TUE-AM-025) to know more!

[Zhen Wang @zhenwangwz]

Implicit 3D Surfaces are Amazing 😀but Recovery of Fine Detail is Tricky ☹️.

The key idea of our #CVPR2023 paper ALTO (https://t.co/X1w2lOeDSq) is to use an “alternating block”, which recovers more detail and also improves runtime over point latents.

📅Poster: TUE-AM-025👋👋👋

[Zhen Wang @zhenwangwz]

(4/n) Cross-dataset evaluation of ALTO and baselines by training on Synthetic Rooms and testing on real-world ScanNet-v2:

[Zhen Wang @zhenwangwz]

Please consider dropping by our poster at 182b this afternoon! Come and say hi! #CVPR2022

[Zhen Wang @zhenwangwz]

We’re presenting our work “Synthetic Generation of Face Videos with Plethysmography Physiology” at #CVPR2022. We study how to achieve physio-realism on the generated face videos and show how this can help #rPPG medical application.
#DeepFakeforSocialGood #HealthTech #fairness

[Zhen Wang @zhenwangwz]

Project webpage: https://t.co/4OMoJnVq6n.

Joint work with @YunhaoBa @PradyumnaChari @DrLalehJ @AchutaKadambi and talented UCLA undergraduates and masters.

We’re going to CVPR in person. Come and say hello (Poster 4.2)!

[Zhen Wang @zhenwangwz]

We’re presenting our work “Synthetic Generation of Face Videos with Plethysmography Physiology” at #CVPR2022. We study how to achieve physio-realism on the generated face videos and show how this can help #rPPG medical application.
#DeepFakeforSocialGood #HealthTech #fairness

[Zhen Wang @zhenwangwz]

1⃣ We propose a scalable biophysics based rendering method to generate synthetic rPPG face videos given any input reference image and target rPPG signals.

2⃣ We also release the largest real rPPG dataset with diverse skin tones to benchmark heart rate estimation performance.

[Achuta Kadambi @AchutaKadambi]

CVPR's 10th Annual comp imaging workshop is co-hosted by our lab today. We cannot control that it falls on Father's Day Sunday, but thank our amazing speakers who carved out time incl @Jimantha @KostasPenn Antonio Torralba, Anat Levin, @debfx, @AB2World, and others!