Pier Luigi Dovesi

Bologna - Shanghai - Stockholm

[Pier Luigi Dovesi @dovpie]

🚀 Daily, we marvel at #AI progress, yet one human ability stands unchallenged: immediate adaptation.

💡 Today, we're closing that gap! Introducing 💀 To Adapt or Not to Adapt? Real-Time Adaptation for Semantic Segmentation. An AI that shifts gears as swiftly as we do! #ICCV2023

[Liner @search_liner]

Liner is partnering with @spoticlr at #ICLR2026 — supporting Best Paper and Travel Awards for LLM research.

And to celebrate, we're giving away:
✈️ Round-trip flights + hotel to #ICML2026 in Seoul
🎁 $300 Liner Credits

Follow @search_liner + repost to enter by 4/27.

Liner is built for research workflows. Find papers, verify sources, and write with citations in one place.

See you in 🇧🇷 and 🇰🇷!

@iclr_conf @icmlconf

[AI at AMD @AIatAMD]

From stereo vision to LiDAR fusion, and the edge cases that break them.

Join the next AMD AI Research Club session where Pier Luigi Dovesi, Robotics and Autonomous Systems Lead, AMD Silo AI, explores the evolving landscape of depth foundation models with Luca Bartolomei.

📅 April 16, 2026
⏰ 4 PM CET | 10 AM CT

We’ll unpack:
• Virtual pattern projection
• Cross-domain depth generalization
• Geometry-aware perception in complex environments

Built for developers, researchers, and engineers pushing the boundaries of AI perception.

👉 Register here: https://t.co/JWkWQvkO8p

[TUM Computer Vision Group @tumcvg]

We are thrilled that our group has twelve papers accepted at #CVPR2025! 🚀
Congratulations to all of our students for this great achievement! 🎉

For more details, check out: https://t.co/pTTuMlMUdp

[Linus Härenstam-Nielsen @LinusHNielsen]

We are presenting DiffCD tomorrow morning at #ECCV2024!

Come by poster 214 to discuss how to fit neural SDFs to point clouds, from a geometric perspective

[Matteo Poggi @mattpoggi]

🍸🍸The TRICKY 24 challenge on "Monocular Depth from Images of Specular and Transparent SurfacesHR Depth from Images of Specular and Transparent Surfaces" is live! 🍸🍸

More details about the challenge: https://t.co/6F6SoQe337

and the TRICKY workshop: https://t.co/cOmyFO5cED

[Matteo Poggi @mattpoggi]

Working on NeRF/3DGS-SLAM for #ECCV2024?
Maybe our survey can help you to write down the tables 😅

[@LuciaGoracci @ZiaLulli1]

Era questione di ore. La notizia è arrivata. Se sei ferito senza che i soccorsi siano consentiti, questo è il tuo destino. Come è destino che l’ennesima morte di un reporter ucciso mentre faceva il suo lavoro, non muterà il corso delle cose.
Noi, oggi, piangiamo un collega
#Gaza

[Pier Luigi Dovesi @dovpie]

Diary of a Genocide. How complicit silence erodes the very fabric of our humanity. https://t.co/jwf8OOQaE6

[Amnesty International @amnesty]

CEASEFIRE NOW!
CEASEFIRE NOW!
CEASEFIRE NOW!
CEASEFIRE NOW!

In the face of unfettered devastation and suffering, humanity must prevail. Demand a ceasefire by all parties to end civilian suffering.

Act now ⬇️
https://t.co/66L4Pe7D8I

[Pier Luigi Dovesi @dovpie]

Check out our poster today at ICCV23!

[AK @_akhaliq]

To Adapt or Not to Adapt? Real-Time Adaptation for Semantic Segmentation

paper page: https://t.co/qWaVPMlvK7

The goal of Online Domain Adaptation for semantic segmentation is to handle unforeseeable domain changes that occur during deployment, like sudden weather events. However, the high computational costs associated with brute-force adaptation make this paradigm unfeasible for real-world applications. In this paper we propose HAMLET, a Hardware-Aware Modular Least Expensive Training framework for real-time domain adaptation. Our approach includes a hardware-aware back-propagation orchestration agent (HAMT) and a dedicated domain-shift detector that enables active control over when and how the model is adapted (LT). Thanks to these advancements, our approach is capable of performing semantic segmentation while simultaneously adapting at more than 29FPS on a single consumer-grade GPU. Our framework's encouraging accuracy and speed trade-off is demonstrated on OnDA and SHIFT benchmarks through experimental results.

[Yann LeCun @ylecun]

Selfied after talking with students in front of the Meta booth at ICCV - Paris

[AK @_akhaliq]

ICCV org on @huggingface crossed 100 team members and 149 papers have been claimed by authors

ICCV page: https://t.co/RWQ7tcyJoN
org: https://t.co/QVrqc86PV1
demo: https://t.co/1CJ0AdeBrX

[TUM Computer Vision Group @tumcvg]

We are happy to announce that 4 papers were accepted to #ICCV2023! Here is a short overview of all the amazing work our team and collaborators have done over the last months. See you at #ICCV2023 in Paris in October !🇫🇷✈️ 1/n

[TUM Computer Vision Group @tumcvg]

Hardware-Aware Modular Least Expensive Training (HAMLET) is a real-time domain adaptation framework for semantic segmentation.
✅ 34% less back-propagation FLOPS
✅ 29FPS on consumer-grade GPUs
9/n

[AK @_akhaliq]

https://t.co/4n2b8GGk0k email for 28th July is out

[Pier Luigi Dovesi @dovpie]

@_akhaliq 💀 Project page: https://t.co/a1D9SzDWbN
📜 Arxiv: https://t.co/rdEOu5Prdr
🧑🏻‍💻 Code: https://t.co/9LG6qVMwrF
📽️ Video: https://t.co/9EK4vdoMJJ

[AK @_akhaliq]

To Adapt or Not to Adapt? Real-Time Adaptation for Semantic Segmentation

paper page: https://t.co/qWaVPMlvK7

The goal of Online Domain Adaptation for semantic segmentation is to handle unforeseeable domain changes that occur during deployment, like sudden weather events. However, the high computational costs associated with brute-force adaptation make this paradigm unfeasible for real-world applications. In this paper we propose HAMLET, a Hardware-Aware Modular Least Expensive Training framework for real-time domain adaptation. Our approach includes a hardware-aware back-propagation orchestration agent (HAMT) and a dedicated domain-shift detector that enables active control over when and how the model is adapted (LT). Thanks to these advancements, our approach is capable of performing semantic segmentation while simultaneously adapting at more than 29FPS on a single consumer-grade GPU. Our framework's encouraging accuracy and speed trade-off is demonstrated on OnDA and SHIFT benchmarks through experimental results.

[Pier Luigi Dovesi @dovpie]

🌱 This work is a collaboration between Marc Botet Colomer*, myself*, Theodor Panagiotakopoulos, João Frederico Carvalho, @LinusHNielsen, @HosseinAzizpour, Hedvig Kjellström, Daniel Cremers, and @mattpoggi .

[Pier Luigi Dovesi @dovpie]

🚀 Daily, we marvel at #AI progress, yet one human ability stands unchallenged: immediate adaptation.

💡 Today, we're closing that gap! Introducing 💀 To Adapt or Not to Adapt? Real-Time Adaptation for Semantic Segmentation. An AI that shifts gears as swiftly as we do! #ICCV2023

[Pier Luigi Dovesi @dovpie]

💀 Project page: https://t.co/a1D9SzDWbN
📜 Paper: https://t.co/rdEOu5Prdr
🧑🏻‍💻 Code: https://t.co/9LG6qVMwrF