@AlessandroFlabo

Co-founder of Procedo | PhD in Computer Science @SapienzaRoma | Member of the Perception and Intelligence Lab. @SapienzaPINlab

 CTO and Founder of Procedo
procederai.com
Joined November 2012
  • Tweets3.8K
  • Following518
  • Followers367
402Media
Include

Only show posts containing:

Exclude

Hide posts containing:

Time range
-
Minimum likes
Replying to @_italai
Loved sharing time and discussing the future of AI with the Panasonic and ItalAI teams! 🔥
2
50
Alessandro Flaborea retweeted
Pascal Mettes@PascalMettes4 Mar 2025
A couple of months ago, I told you vision-language = hyperbolic. #ICLR2025 agrees and has made our latest paper an oral presentation! To present at ICLR: Compositional Entailment Learning for Hyperbolic Vision-Language Models. Link: arxiv.org/abs/2410.06912

Compositional Entailment Learning for Hyperbolic Vision-Language Models

Image-text representation learning forms a cornerstone in vision-language models, where pairs of images and textual descriptions are contrastively aligned in a shared embedding space. Since visual...

arxiv.org
Pascal Mettes@PascalMettes18 Oct 2024
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:
2
19
133
9,450
Alessandro Flaborea retweeted
Fabio Galasso@GalassoFab1011 Feb 2025
"Compositional Entailment Learning for Hyperbolic Vision-Language Models" #ICLR25 has been 𝐬𝐞𝐥𝐞𝐜𝐭𝐞𝐝 𝐟𝐨𝐫 𝐚𝐧 𝐨𝐫𝐚𝐥 𝐩𝐫𝐞𝐬𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 🏆 1.8% of the submitted papers were selected for an oral
Fabio Galasso@GalassoFab1025 Jan 2025
"Compositional Entailment Learning for Hyperbolic Vision-Language Models" #ICLR25 On introducing text and image hierarchies in multi-modal learning with hyperbolic neural networks arxiv.org/abs/2410.06912

Compositional Entailment Learning for Hyperbolic Vision-Language Models

Image-text representation learning forms a cornerstone in vision-language models, where pairs of images and textual descriptions are contrastively aligned in a shared embedding space. Since visual...

arxiv.org
2
14
302
Just a quick update on our HyCoCLIP paper: accepted to the #ICLR2025 conference with a avg score of 8! A great collaboration between @SapienzaPINlab (@GalassoFab10 , @GuidoDamely) and @PascalMettes @theAvikPal Paper: arxiv.org/abs/2410.06912 See the thread for more insights 👇🏻

Compositional Entailment Learning for Hyperbolic Vision-Language Models

Image-text representation learning forms a cornerstone in vision-language models, where pairs of images and textual descriptions are contrastively aligned in a shared embedding space. Since visual...

arxiv.org
Avik Pal@theAvikPal18 Oct 2024
(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.
1
1
1
14
690
🔨 AI in manufacturing AI and Computer Vision are revolutionizing Manufacturing in different aspects, from security to predictive maintenance and supply-chain optimization. Is your company still thinking if AI will be considered an integral part of the industry? It will. Soon.
This quoted post is unavailable.
Try viewing anyway · View on X
2
173
Proud to share that this work results from a great collaboration between @SapienzaPINlab and @anfurnari @GMFarinella. Huge thanks to the team for their dedication and drive! 🚀 @GalassoFab10 @LeonardoPlini @lu_scofano @edmatteis @GuidoDamely
1
1
314
⚙️Experiments on two procedural datasets, Assembly101-O and Epic-tent-O, show the power of this approach in dynamic, real-time scenarios.
1
75
💡Our model combines step recognition and future step anticipation using LLMs (#LLAMA, #Phi, #Gemma, #Mistral) with Chain of Thought and In-context learning. Errors are flagged when recognized actions don’t align with anticipated ones, helping detect unforeseen mistakes.
1
45
👀Detecting procedural mistakes online in egocentric videos is critical for industries like manufacturing and healthcare. Many of these mistakes are open-set, meaning they’re unexpected. Our dual-branch approach tackles this by enabling online error detection.
1
50
Excited to share our latest research, TI-PREGO—a follow-up to PREGO, our #CVPR2024 work! 📈 We explore how Chain of Thought and In-context learning can boost mistake detection in procedural videos. Paper: arxiv.org/abs/2411.02570 A good read before the #CVPR2025 deadline! 🚀

TI-PREGO: Chain of Thought and In-Context Learning for Online...

Identifying procedural errors online from egocentric videos is a critical yet challenging task across various domains, including manufacturing, healthcare, and skill-based training. The nature of...

arxiv.org
1
4
13
716
Alessandro Flaborea retweeted
Pushmeet Kohli
@pushmeet11 Nov 2024
The #AlphaFold 3 model code and weights are now available for academic use. We @GoogleDeepMind are excited to see how the research community continues to use AlphaFold to address open questions in biology and new lines of research. github.com/google-deepmind/a…

GitHub - google-deepmind/alphafold3: AlphaFold 3 inference pipeline.

AlphaFold 3 inference pipeline. Contribute to google-deepmind/alphafold3 development by creating an account on GitHub.

github.com
Pushmeet Kohli
@pushmeet13 May 2024
We love the excitement & results from the community on AlphaFold 3 and are doubling the AF Server daily job limit to 20. Happy to also share that we're working on releasing the AF3 model (incl weights) for academic use, which doesn’t depend on our research infra, within 6 months.
21
254
31
1,028
187,149
🚀 Exploring the Research Behind @proceder_ai ! Our latest #CVPR2024 publication introduced a method for spotting anomalies in AR/VR-like procedural videos. This could be a game-changer for real-time applications! So, why not aim to make this tool an everyday asset? 🌟
This quoted post is unavailable.
Try viewing anyway · View on X
1
4
287
Alessandro Flaborea retweeted
Fabio Galasso@GalassoFab1021 Oct 2024
ARIA glasses have arrived today at @SapienzaPINlab Thanks @meta_aria and @AIatMeta for getting us onboard!
4
1
28
1,450
During my PhD, I always wanted to turn my research into something real. Now, I’m on this entrepreneurial journey, eager to learn as much as I can! 😉
This quoted post is unavailable.
Try viewing anyway · View on X
1
10
212
Another great collaboration between @SapienzaPINlab and the University of Amsterdam! Here we propose a novel view on how Hyperbolic Neural Networks represent the multimodal space of images and texts leveraging their intrinsic hierarchical structure.
Avik Pal@theAvikPal18 Oct 2024
(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.
7
239
Choosing a company name is never easy, but in this case, I can confidently say that the blend of my language and our AI-driven vision is a perfect match. #ProcederAI
This quoted post is unavailable.
Try viewing anyway · View on X
1
5
253
Here's the post with links to the paper and code: xitter.cf/AlessandroFlabo/status…
Alessandro Flaborea@AlessandroFlabo9 Sep 2024
🆕News about our oral work at #IROS2024 The final version of the paper is available! arxiv.org/abs/2407.13567 We've released the code, check it (and leave a⭐️): github.com/GDam90/hyp2nav See you in Abu Dhabi! @GuidoDamely @PascalMettes @GalassoFab10
3
130
Today #IROS2024 starts and @SapienzaPINlab is present ✅ @GuidoDamely will have the oral presentation of Hyp2Nav, work done while we were visiting @PascalMettes, on the 18th. He'll show how hyperbolic NNs improve explainability and reduce dimensionality for RL. @GalassoFab10
1
3
1
11
1,214
🔥 @proceder_ai is here! 🔥 I'm happy to share that we're creating Procedo, the AI assistant that can guide and assist you while you're doing everyday complex tasks. Are you ready to proceed with us?
This quoted post is unavailable.
Try viewing anyway · View on X
1
4
109
It's nearly impossible to find a field untouched by the impact of deep learning, and we're seeing the natural consequences of this process. Deep Learning, and AI in general, needs to be seen as a tool, not as an enemy.
Yam Peleg
@Yampeleg9 Oct 2024
Deep Learning. ✅ Turing award. ✅ Physics nobel prize. ✅ Chemistry nobel prize. ☐ Medicine nobel prize. ☐ Literature nobel prize. ☐ Economics nobel prize. ☐ Peace nobel prize. ☐ Fields medal. Gradient. Descent.
3
106
Load more