Ananta R. Bhattarai

PhD Student @ Visual AI for Extended Reality with Prof. Helge Rhodin

[Paul-Edouard Sarlin @pesarlin]

Where is the code? No code, no public API, private datasets, too little details to reproduce = low impact. What does the award committee value?

[Ananta R. Bhattarai @anantarbb]

🧵(5/5)

Project Page: https://t.co/gUfGhf6oNB
Code: https://t.co/H8fAbW5oX6

[Ananta R. Bhattarai @anantarbb]

📢📢Excited to share that Re-Depth Anything: Test-Time Depth Refinement via Self-Supervised Re-lighting will be presented at CVPR 2026 (Findings)!

💡Re-Depth refines monocular depth without any ground-truth labels — by re-lighting the prediction.

🧵(1/n)
#CVPR2026

[Ananta R. Bhattarai @anantarbb]

🧵(4/n)

Results: Applied on top of Depth Anything V2 (DA-V2) and Depth Anything 3 (DA3), Re-Depth improves results both qualitatively and quantitatively — sharpening fine details and cleaning noise off flat surfaces.

[Artem Sevastopolsky @ASevastopolsky]

❌ Tracking by 68 sparse face landmarks?
✅ Tracking by dense, per-pixel head landmarks

📣 DenseMarks: Learning Canonical Embeddings for Human Heads Images via Point Tracks

https://t.co/GBKBUYHaSb

🧵

[Helge Rhodin @HelgeRhodin]

Turns out Shape from Texture is useful after all!
Our DreamTexture method brings this ancient technique to the age of generative models. We reconstruct 3D by synthesizing virtual textures and augmenting them on real images:
https://t.co/AbsWZ1g4E8
⚽️🦓🐝🧊

[Ananta R. Bhattarai @anantarbb]

Big thanks to my friends at the Visual Computing Lab for their help and insightful discussion.

[Ananta R. Bhattarai @anantarbb]

I am extremely excited to share that our paper, TriPlaneNet, has been accepted to WACV 2024. Thank you @ASevastopolsky and @MattNiessner for providing me with an incredible opportunity to work on such an exciting project.

[Matthias Niessner @MattNiessner]

Check out TriPlaneNet!

From a single image, we predict EG3D latents & offsets, thus obtaining high-fidelity 3D models. Works in real time!

Great work by our stellar MA student @anantarbb advised by @ASevastopolsky #WACV2024

https://t.co/0PNl26iaDW
https://t.co/45h21TF27N

[Matthias Niessner @MattNiessner]

Check out TriPlaneNet!

From a single image, we predict EG3D latents & offsets, thus obtaining high-fidelity 3D models. Works in real time!

Great work by our stellar MA student @anantarbb advised by @ASevastopolsky #WACV2024

https://t.co/0PNl26iaDW
https://t.co/45h21TF27N