TeV - Technologies of Vision @FBK

Part of @FBK_research, pioneering #ComputerVision #AI & #DeepLearning since 1986. Advancing research & industrial applications with a system-level approach.

[IEEE Access @IEEEAccess]

Edge clarity is vital for interpreting visual information, but poor lighting, motion, or low-quality capture can make details disappear. This featured IEEE Access article introduces a new image-aware sharpening filter: https://t.co/qYhLNpxlvh

[Wired Italia @wireditalia]

Minuscole fibrille, intelligenza artificiale e un QR code che non si lascia copiare: così l'Ipzs e la Fondazione Bruno Kessler hanno creato un’etichetta capace di provare la propria autenticità https://t.co/XKrnqGHVjx

[PAVIS IIT @PAVIS_IIT]

Today Matteo Bortolon defending his PhD work done in collaboration with @TeV_FBK, PAVIS and @UniTrento on Neural Representations for Robotics. Congratulations Doctor! 🎓

[TeV - Technologies of Vision @FBK @TeV_FBK]

Can you spot our TeV00 Robot?

We are warming up for three days of non-stop demos here at Wired Next Festival in Rovereto!

If you are around, come visit us at the Fondazione Caritro (Fondazione VRT) booth in Corso Bettini and discover how TeV00 can listen, reason, and grasp.

[TeV - Technologies of Vision @FBK @TeV_FBK]

A week full of field experiences! 📷

✅Raspberry grading at Sant’Orsola passed the test – thousands sorted in real time with #MotionCam3D🫐

✅Olive tree digital twins built with #ProjectAria + #Insta360 🌿

#AgileHand #AgRimate #AgriTech
@FBK_research

[Fondazione Bruno Kessler - FBK @FBK_research]

FBK a Wired Next Fest Trentino: il festival dedicato a innovazione, tecnologia, scienza e cultura ospiterà diverse talk di FBK.
📍Vi aspettiamo a Rovereto dal 2 al 5 ottobre!

🔎Info e programma (potrebbe subire variazioni): https://t.co/hqrp4fJ4BC

[TeV - Technologies of Vision @FBK @TeV_FBK]

la @NotteRicercaTN é cominciata alla grande, tev00 ha subito catturato l'attenzione @FBK_research #AI #Robotics

[TeV - Technologies of Vision @FBK @TeV_FBK]

Lots of valuable feedback & engaging exchanges with fellow researchers — a great opportunity to share, learn, and grow 🚀

#AI #ComputerVision #VisualMonitoring #ResearchCommunity #AI4Industry #TrustTech

[TeV - Technologies of Vision @FBK @TeV_FBK]

✅ Last week went very well for Davide Boscaini at #ICIAP2025 in Rome!
He presented at the BioFor 2025 Workshop and showcased our work during the poster session.
🧵👇

[TeV - Technologies of Vision @FBK @TeV_FBK]

🔹 AI-driven Visual Monitoring of Industrial Assembly Tasks (ViMAT)
ViMAT is a novel AI system for real-time monitoring of assembly tasks.
No rigid setups or markers needed.
Perception + probabilistic reasoning → safe, flexible, and accurate monitoring in dynamic environments.

[TeV - Technologies of Vision @FBK @TeV_FBK]

👥Authors:
@weijler_lisa, @sebastiankoch98, @Poiex, Timo Ropinski, @Pedro_He_Ca | TU Wien · Ulm Univ · @FBK_research

🔗 Project page: https://t.co/lVWYXXGKm2

#NeurIPS2025 #AI #ComputerVision #3D #NeRF #RepresentationLearning

[TeV - Technologies of Vision @FBK @TeV_FBK]

🎉Exciting news! The paper “OpenHype: Hyperbolic Embeddings for Hierarchical Open-Vocabulary Radiance Fields” has been accepted at #NeurIPS2025

3D objects and scenes naturally have hierarchical structures. But current approaches struggle with multi-level granularity.

[TeV - Technologies of Vision @FBK @TeV_FBK]

🔎 OpenHype introduces a new solution: representing scene hierarchies in a continuous hyperbolic latent space.

This enables:
🔹 Natural encoding of multi-scale relationships
🔹 Smooth traversal of hierarchies via geodesic paths
🔹 Superior segmentation performance & efficiency

[TeV - Technologies of Vision @FBK @TeV_FBK]

We’re getting ready for the European Researchers’ Night 2025!
📍Sept 26 | @MUSE_Trento | from 5PM
Find us at floor +1, booth 26

🤖 TEV00 – The robotic arm that listens, thinks & grasps: from text commands to intelligent actions, live in action!
👉https://t.co/EkxHAUx5jU
#AI #FBK

[TeV - Technologies of Vision @FBK @TeV_FBK]

🚀Our colleague Davide Boscaini is in Rome for @iciapconf

🗓️Fri, Sept 19 | 14:00–15:30 | Poster Session 3
📍Rectorate Building, @SapienzaRoma

He’ll present: AI-driven Visual Monitoring of Industrial Assembly Tasks (ViMAT)

🔗 https://t.co/6aQIxorsjG

#AI #CV @FBK_research

[TeV - Technologies of Vision @FBK @TeV_FBK]

🎉 Proud to celebrate our colleague @sergiopovoli, now one of FBK’s Science Ambassadors!
Excited to keep sharing research and science with him 🌍✨
🔗 https://t.co/ogFZs4lO3S

#FBK #ScienceAmbassadors #ScienceCommunication

[Fondazione Bruno Kessler - FBK @FBK_research]

Think loud, speak science: le voci della ricerca scientifica
Si occupano di sostenibilità, tech quantistiche, micro-nanofabbricazione, diversità religiose e molto altro: sono i 24 nuovi Science Ambassador di FBK, pronti a raccontare ricerca e innovazione

https://t.co/3k6P24foFb

[TeV - Technologies of Vision @FBK @TeV_FBK]

🎉 Two papers related to the FishGrade dataset have been accepted in IEEE/CAA Journal of Automatica Sinica!

📄Vision-Guided Robotic System for Fish Quality Grading & Packaging
📄The AGILEHAND architecture

@FBK_research @AgileHand_

#AgriTech #ComputerVision #AI4Seafood

[TeV - Technologies of Vision @FBK @TeV_FBK]

🚀New dataset on Hugging Face!

FishGrade🐟for real-time, non-invasive grading of fish steaks using RGB+depth and deep learning.

Annotations for segmentation + quality classes.

By @TeV_FBK, part of #AgileHandEU🇪🇺

🔗https://t.co/YUaCPDRlgo

#AgriTech #AI4Seafood #ComputerVision

[TeV - Technologies of Vision @FBK @TeV_FBK]

💡The result? Generalized & transferable 3D representations with strong performance on downstream tasks.

Paper: https://t.co/oQHuNtfpjW
Code: https://t.co/FLOB8Ln7vy

Abiao Li, @Aliexken, Yuming Fang, Yifan Zuo, Jian Zhang, @CvGfmei
@FBK_research

#ICCV2025 #3DVision #DL

[TeV - Technologies of Vision @FBK @TeV_FBK]

🚀 Our paper “PointGAC: Geometry-Aware Clustering for Masked Point Cloud Modeling” will be presented at ICCV 2025! 🎉

🔎Why PointGAC matters: Most masked point cloud modeling (MPM) methods rely on regression, often overfitting to local details.

[TeV - Technologies of Vision @FBK @TeV_FBK]

PointGAC uses a clustering-based teacher–student framework to:
📐Partition patches geometry-aware
🌐Align masked features with cluster centers
⚡Maintain an adaptive k-means codebook