Guillem Brasó
@GuillemBraso
Followers
190
Following
716
Media
4
Statuses
33
Research Scientist @NVIDIA
San Francisco, CA
Joined December 2017
Excited to share what we've been working on! SeNaTra introduces a backbone where segmentation emerges natively by replacing standard downsampling with grouping layers. Opens the door for a new family of zero-shot segmentation-centric backbone architectures! 🚀.Code coming soon!
The time for new architectures is over? Not quite! SeNaTra, a native segmentation backbone, is waiting, let's see how it works 🧵
2
15
98
RT @huangjh_hjh: [1/N] 🎥 We've made available a powerful spatial AI tool named ViPE: Video Pose Engine, to recover camera motion, intrinsic….
0
91
0
RT @HuanLing6: We are excited to share Cosmos-Drive-Dreams 🚀 .A bold new synthetic data generation (SDG) pipeline powered by world foundati….
0
43
0
RT @AljosaOsep: Turns out that if you learn to downsample (rather than using uniform grid pooling) in Vision Transformers, you no longer ne….
arxiv.org
Uniform downsampling remains the de facto standard for reducing spatial resolution in vision backbones. In this work, we propose an alternative design built around a content-aware spatial grouping...
0
9
0
RT @xuanchi13: 📢🚗✨ Excited to announce InfiniCube, our scalable generative model for dynamic 3D driving scene generation with high fidelity….
0
47
0
RT @JennySeidensch1: Had such a great time at @CVPR presenting our GHOST 👻 tracker at the tracking and it's many guises workshop (thanks @A….
0
1
0
RT @lealtaixe: Multi-object tracking methods often feel hacky. It can be frustrating to read a nice paper, only to find a series of "post-p….
0
37
0
Check out our new work: a unified solution for short and long-term multi-object tracking using graph hierarchies!🦾🦾.Paper: Code coming soon!.@OrcunCetintas @lealtaixe.
arxiv.org
Tracking objects over long videos effectively means solving a spectrum of problems, from short-term association for un-occluded objects to long-term association for objects that are occluded and...
Multi-object tracking methods often feel hacky. It can be frustrating to read a nice paper, only to find a series of "post-processing" hacks to improve the numbers. It is as if the tracking problem could not be solved in a single step. Are you feeling this too? Let's dig👇.
1
1
10
RT @Peter4AI: We are gonna present our work on low data generalization at #NeurIPS2022 tomorrow (28.11, Hall J.125)! .Happy to chat and dis….
0
20
0
RT @lealtaixe: After the success of MPNTrack, we push the boundaries of instance segmentation of multiple objects! Paper: .
github.com
[IJCV 2022] Multi-Object Tracking and Segmentation via Neural Message Passing - ocetintas/MPNTrackSeg
0
6
0
RT @Peter4AI: Happy to share that our paper got accepted to #NeurIPS2022! 🥳. If you are working on Computer Vision with low amount of data,….
0
57
0
Incredibly proud of my advisor and the whole team!❤️.
Always good to receive good news during vacations: we got the @ERC_Research Starting Grant! I say "we" because, yes, it is an individual grant, and no, it cannot possibly be just an individual grant 🧵 (0/7).
0
0
8
RT @JennySeidensch1: After an awesome first in-person conference three weeks ago, this week I attend the international computer vision summ….
0
4
0
Announcing the 7th edition of the BMTT workshop at #CVPR2022, focusing on how far synthetic data can take us in tracking and beyond🚀. Check out our #syn2real challenges and speakers line-up!. Site: @Matteo18025562 @AljosaOsep @OrcunCetintas @lealtaixe
1
8
30
RT @lealtaixe: To all CV/ML Ukrainian PhD students, if you are interested in our research (, I am offering paid int….
0
14
0
RT @lealtaixe: Syn2real arrives for MOT! Excited to release MOTSynth to study the use of synthetic data for detection, tracking, and reid.….
0
24
0
RT @Matteo18025562: We are releasing MOTSynth, the largest dataset for Pedestrian Detection and Tracking. Expect to obtain better real-wor….
0
81
0
RT @cristiancanton: Always find some time to mentor the raising starts who will become the top researchers of the future! Thanks for your q….
0
3
0
RT @_M_Weber: Our paper STEP: Segmenting and Tracking Every Pixel is presented at #NeurIPS2021 Track on Datasets and Benchmarks. We're prop….
0
9
0