News Archive

Reconstructing 3D motion from a single camera is usually a mathematical nightmare, but Svitlana built MOSAIC-GS to make it surprisingly fast and high-quality by basically pre-calculating the chaos.

I will serve as Area Chair for ECCV 2026!

Since ‘the’ and ‘and’ are apparently just giant attention-hungry magnets, Anna figured out how to filter the noise and turn them into RefAM, a zero-shot segmenter that somehow hits SOTA without us having to train a single thing.

Eruku got accepted to WACV2026, congrats to all my co-authors! In this work we improve on the handwriting generation capabilities of our previous work Emuru by making the model more robust and flexible in its usage.

Congratulations to Luca Zanella for having our joint work on online video step grounding: BagLM presented at Neurips2025.

Gemini 2.5 report is out! Proud to see me and my team in the contributors list.

Congratulations to Enis for having UIP2P accepted at ICCV25!

CVPR 2025: Presenting 4 works: EgoTempo Benchmark, ACED, VICT, and Emuru. Honored to be recognized as outstanding reviewer!

Another of our CVPR works is now online, check our EgoTempo Benchmark if you working on Egocentric long form video understanding. A huge thanks to Chiara Plizzari for the amazing work she did while she was a student in our team at Google Zurich!

4 works have been accepted to CVPR, 1 is already online ACED, 3 more to come 👀

WACV 2025: LIME will be presented as oral! Congratulation to Enis.

New papers on Arxiv: UIP2P (unsupervised image editing) and ACED (data curation for small VLM).

ECCV 2024: 2 papers on building better VLM accepted! Check TCR and BRAVE.

TouchSDF accepted in RAL! Check also the follow-up SnapTapSplat on Arxiv.

Check our latest works on Arxiv: Lime, InseRF, and TCR.

3DV: Our TextMesh work has been accepted!

ICCV 2023: NeRF-GAN Distillation and LatentSwap3D presented at the AI3DCC workshop.

CVPR 2023: NeRF-supervised Deep Stereo accepted. See you in Vancouver!

AT/DT++ has been accepted for publication on TPAMI.

Honored to be recognized as **Outstanding Reviewer at ECCV 2022!

Interviewed for the ProfessionAI podcast sharing experience as a CV researcher (link).

Uploading new works to Arxiv: LatentSwap3D and ParGAN.

BNE accepted for publication on Pattern Recognition.

CVPR 2021: BNE presented at the L2ID workshop. Honored to be Outstanding Reviewer at CVPR 2021!

LegoFormer uploaded to Arxiv with open source code.

Extended version of our CVPR 2019 oral accepted to TPAMI (realTimeII).

3DV 2020: Divide et Impera accepted as an Oral presentation!

Recognized as Outstanding Reviewer at ECCV 2020 and ACCV 2020.

BNE and Continual Online Adaptation for Deep Stereo (TPAMI sub) uploaded to Arxiv.

AdaptationJournal accepted for publication on the RGBD special issue of TPAMI.

Semi-Automatic Labeling for Deep Learning in Robotics accepted for publication in IEEE TASE.

ICCV 2019: Learning Across Tasks and Domains (ATDT) accepted for poster presentation.

Started collaboration with Federico Tombari’s team at Google Zurich. So long Bologna, and thank you for all the fish!

CVPR 2019: Presenting Real-time self-adaptive deep Stereo (Oral) and Learning To Adapt for Stereo (Poster).

Successfully defended my Ph.D. Thesis: Computer Vision and Deep Learning for Retail Store Management.

DIHE published on CVIU and FPGA Dense Depth published on IEEE TCAS II.