Pascal Mettes
@PascalMettes
Followers
2K
Following
844
Statuses
369
Assistant Professor - University of Amsterdam | Hyperbolic deep learning
Joined March 2020
Our survey "Hyperbolic Deep Learning in Computer Vision: A Survey" has been accepted to #IJCV! The survey provides an organization of supervised and unsupervised hyperbolic literature. Online now: w/ @GhadimiAtigMina @mkellerressel @jeffhygu
@syeung10
2
25
140
0
0
1
@Maze_s_Center @tensorqt In theory this makes a lot of sense, due to the geometric nature of language. As with many things in hyperbolic learning though, someone has yet to do it! Sounds like a great paper ready to be written to me.
0
0
3
@PaolaGrossoWork @UvA_Science @UvA_IvI @NWONieuws @SennayGhebreab @MnsUva @Ciena @ChrysaPapagian1 @illc_amsterdam @DESIRE6G_EU @QuSoft @Supercomputing @WiroNiessen In my view, the only way to make exam clarifications/corrections scalable and pleasant is with a fixed window to drop by. Faster (at least for me, slow typer) and makes students feel more involved.
0
0
0
2
0
0
@BlackHC The conflict of interest part is an interesting one in the general sense, as this is an intimate field. People know each other, or work together. Not sure how big of an issue this is, I'm not high up enough in the food chain to have seen insights or numbers :P
1
1
4
@BlackHC I don't know what this means, nor do I see us as any different. I wouldn't even know how this helps me in any way. All I want is that we stop calling our colleagues names.
1
0
0
@BlackHC Look, we all have papers that are not attributed as we would like (or much worse). We all make papers that cite imperfectly as well. I don't have a stake in this. To me, this is more about ego than about the state of academia.
1
0
0
RT @eTalaveraM: I have an open PhD position on Multimodal Learning and Computer Vision. It will be under my supervision! 💫 The position i…
0
91
0
@plain_simon We started like that indeed, but the implications of hyperbolic learning are far bigger. E.g., vision-language models greatly benefit from hyperbolic embeddings, OOD improves, molecular generation is better, etc. etc. Every day we find new potential and practical use-cases.
0
0
5
@ponguru @c_voelcker @VishnuBoddeti @ElaheArani @peterbhase @weakmath @joisino_en @MartinPawelczyk @S__Schoepf @NailaMurray @RaiaHadsell @PaulVicol @PierreAblin @adiyossLC @Azizzadenesheli @VictorKnox99 @priyanshul1202 @koiRitwikHai @bhaskarahanuma9 @akshitwt @vnnm404 @AlamuruRuthwik @melamparithy @surya71704 @swaroop44014330 Very cool, will check it out further!
0
0
2
RT @_italai: Meet "Hyperbolic Active Learning for Semantic Segmentation under Domain Shift' - a game changer for data labelling 🌐 HALO max…
0
1
0
All vision-language models should have hyperbolic embeddings. Vision and language are incredibly hierarchical in nature! See below our latest work on hyperbolic vision-language models that exploit visual compositions through entailment:
(1/6)🥳 Excited to share my latest research done as part of my MSc AI thesis! We introduced Hyperbolic Compositional CLIP (HyCoCLIP)—a novel framework that leverages the hierarchical nature of hyperbolic space for learning vision-language representations using scene compositions.
2
11
140
RT @AlessandroFlabo: Today #IROS2024 starts and @SapienzaPINlab is present ✅ @GuidoDamely will have the oral presentation of Hyp2Nav, work…
0
3
0
RT @dimtzionas: Reconstructing 3D human-object interactions from images/videos is a key step towards realistic #3D humans for #AR/#VR and a…
0
21
0
RT @rastegar_sarah: 🚀 Excited to present our work on Self-Expertise at #ECCV2024 in Milan! Join us at poster #89 on Friday, Oct 4 at 10:30…
0
8
0