Cadar

I like to match keypoints - PhD Student. Universidade Federal de Minas Gerais, Brazil. Université de Bourgogne, France.

[Cadar @eucadar]

A special command for #CVPR2024 in PaperHound!
just append "c+CVPR2024" to filter the papers ;)

https://t.co/mFJU0ZRImq

[Cadar @eucadar]

weekend project

https://t.co/gTUaUboO5n

[Cadar @eucadar]

@Parskatt @chrisoffner3d @XiaoweiZhou5 @XingyiHe1 @pengsida Looking at the power lines on the left, I think its the same side (but I'm not really sure).

[Cadar @eucadar]

@Parskatt @chrisoffner3d @XiaoweiZhou5 @XingyiHe1 @pengsida Maybe related.
Just saw that with RoMa
Showing inliers and outliers with green/red

[dbrand @dbrand]

Congrats to @LinusTech on the Commuter Backpack launch! Looks great. What an amazing product.

We were in the process of trashing ours, then realized it could be used to farm engagement instead.

RT to win a free Commuter Backpack, shipped anywhere in the world. We'll pick one winner in 24 hours.

[Cadar @eucadar]

First panel on the right side

[Cadar @eucadar]

Hey ACCV folks, tomorrow I���ll be presenting our work on improving image correspondence using semantic cues.

“Leveraging Semantic Cues from Foundation Vision Models for
Enhanced Local Feature Correspondence”

Poster 1.29 :)

#ACCV2024 @ACCVConf

[Du Tran @dutran]

prof. Michal Irani is giving a keynote about "Reading Minds and Machines"

[Cadar @eucadar]

Doppelgangers datasets are cool

[Zhenjun Zhao @zhenjun_zhao]

Doppelgangers++: Improved Visual Disambiguation with Geometric 3D Features

@ambie_kk, @ruojin8, Hanyu Chen, Jeffrey Byrne, @Jimantha

tl;dr: scale training data in Doppelgangers; replace CNN-based classifier in Doppelgangers with pre-trained MASt3R

https://t.co/FJHcZwkkI3

[Nathan Lambert @natolambert]

First NeurIPS social starting in ~30 mins 🥳

[Cadar @eucadar]

📄Arxiv: https://t.co/Xq2l4hlemU
💻Code: https://t.co/lGAVZOKCpi

✨TLDR reviews:
- https://t.co/khbyO5euFY
- https://t.co/A0mshWwtb2

[Zhenjun Zhao @zhenjun_zhao]

Leveraging Semantic Cues from Foundation Vision Models for Enhanced Local Feature Correspondence

@eucadar, @guipotje, @renatojmsdh, Cédric Demonceaux, Erickson R. Nascimento

tl;dr: single view->XFeat+DINOv2->SG-like context aggregation->LG supervision

https://t.co/y5WraKaW7T

[Cadar @eucadar]

Hey ACCV folks, tomorrow I’ll be presenting our work on improving image correspondence using semantic cues.

“Leveraging Semantic Cues from Foundation Vision Models for
Enhanced Local Feature Correspondence”

Poster 1.29 :)

#ACCV2024 @ACCVConf

[Cadar @eucadar]

It is time

https://t.co/G0lVmCk3Av

[Cadar @eucadar]

Now one of many position embeddings

[Cadar @eucadar]

Now let’s add all the “little” details other people figured out for us

Time for registers

[Cadar @eucadar]

@edsonroteia You are absolutely right. Thank you (although it didn’t show improvements, it’s a nice experiment)

[Cadar @eucadar]

me enganei, é de graça até 23h

[Cadar @eucadar]

poxa, show do mc rick ta 20 conto

[Cadar @eucadar]

em bh achei que ele podia subir isso aí

[Cadar @eucadar]

impressora 3d quebrou. Agora preciso de uma segunda impressora pra arrumar a primeira

[Igor Rogh @igorogh]

ABSOLUTE CINEMA

[Cadar @eucadar]

N eh possível man